Redundant aggregation and exchange makes the query much slower

## Bug Report

### 1. Minimal reproduce step (Required)
TPC-H lineitem table.
``` SQL
mysql> show create table lineitem;
+----------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+----------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| lineitem | CREATE TABLE `lineitem` (
`L_ORDERKEY` bigint(20) NOT NULL,
`L_PARTKEY` bigint(20) NOT NULL,
`L_SUPPKEY` bigint(20) NOT NULL,
`L_LINENUMBER` bigint(20) NOT NULL,
`L_QUANTITY` decimal(15,2) NOT NULL,
`L_EXTENDEDPRICE` decimal(15,2) NOT NULL,
`L_DISCOUNT` decimal(15,2) NOT NULL,
`L_TAX` decimal(15,2) NOT NULL,
`L_RETURNFLAG` char(1) NOT NULL,
`L_LINESTATUS` char(1) NOT NULL,
`L_SHIPDATE` date NOT NULL,
`L_COMMITDATE` date NOT NULL,
`L_RECEIPTDATE` date NOT NULL,
`L_SHIPINSTRUCT` char(25) NOT NULL,
`L_SHIPMODE` char(10) NOT NULL,
`L_COMMENT` varchar(44) NOT NULL,
PRIMARY KEY (`L_ORDERKEY`,`L_LINENUMBER`) /*T![clustered_index] NONCLUSTERED */
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin |
+----------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)

mysql> explain analyze select a.L_ORDERKEY, sum(a.L_PARTKEY) from lineitem a left join lineitem b on a.L_ORDERKEY = b.L_LINENUMBER group by a.L_ORDERKEY;
+----------------------------------------------------+--------------+-----------+--------------+---------------+--------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------------+----------+------+
| id | estRows | actRows | task | access object | execution info | operator info | memory | disk |
+----------------------------------------------------+--------------+-----------+--------------+---------------+--------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------------+----------+------+
| Projection_7 | 7623680.00 | 7500000 | root | | time:6.06s, loops:7427, Concurrency:5 | sf10.lineitem.l_orderkey, Column#35 | 485.4 KB | N/A |
| └─TableReader_48 | 7623680.00 | 7500000 | root | | time:6.01s, loops:7427, cop_task: {num: 261, max: 0s, min: 0s, avg: 0s, p95: 0s, copr_cache_hit_ratio: 0.00} | data:ExchangeSender_47 | N/A | N/A |
| └─ExchangeSender_47 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:6s, min:5.72s, p80:6s, p95:6s, iters:768, tasks:3, threads:60} | ExchangeType: PassThrough | N/A | N/A |
| └─Projection_43 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:5.86s, min:5.55s, p80:5.86s, p95:5.86s, iters:768, tasks:3, threads:60} | Column#35, sf10.lineitem.l_orderkey | N/A | N/A |
| └─HashAgg_44 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:5.9s, min:5.58s, p80:5.9s, p95:5.9s, iters:768, tasks:3, threads:3} | group by:sf10.lineitem.l_orderkey, funcs:sum(Column#37)->Column#35, funcs:firstrow(sf10.lineitem.l_orderkey)->sf10.lineitem.l_orderkey | N/A | N/A |
| └─ExchangeReceiver_46 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:2.56s, min:2.53s, p80:2.56s, p95:2.56s, iters:258, tasks:3, threads:60} | | N/A | N/A |
| └─ExchangeSender_45 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:2.58s, min:0s, p80:2.58s, p95:2.58s, iters:768, tasks:3, threads:60} | ExchangeType: HashPartition, Hash Cols: [name: sf10.lineitem.l_orderkey, collate: binary] | N/A | N/A |
| └─HashAgg_12 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:2.47s, min:0s, p80:2.47s, p95:2.47s, iters:768, tasks:3, threads:3} | group by:Column#42, funcs:sum(Column#41)->Column#37 | N/A | N/A |
| └─Projection_52 | 118051610.25 | 137138461 | mpp[tiflash] | | tiflash_task:{proc max:653.5ms, min:0s, p80:653.5ms, p95:653.5ms, iters:2098, tasks:3, threads:60} | cast(sf10.lineitem.l_partkey, decimal(20,0) BINARY)->Column#41, sf10.lineitem.l_orderkey | N/A | N/A |
| └─HashJoin_42 | 118051610.25 | 137138461 | mpp[tiflash] | | tiflash_task:{proc max:713.5ms, min:0s, p80:713.5ms, p95:713.5ms, iters:2098, tasks:3, threads:120} | left outer join, equal:[eq(sf10.lineitem.l_orderkey, sf10.lineitem.l_linenumber)] | N/A | N/A |
| ├─ExchangeReceiver_21(Build) | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:84ms, min:0s, p80:84ms, p95:84ms, iters:1632, tasks:3, threads:60} | | N/A | N/A |
| │ └─ExchangeSender_20 | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:149ms, min:0s, p80:149ms, p95:149ms, iters:574, tasks:3, threads:60} | ExchangeType: HashPartition, Hash Cols: [name: sf10.lineitem.l_orderkey, collate: binary] | N/A | N/A |
| │ └─TableFullScan_19 | 29999795.00 | 29999795 | mpp[tiflash] | table:a | tiflash_task:{proc max:70ms, min:0s, p80:70ms, p95:70ms, iters:574, tasks:3, threads:60} | keep order:false | N/A | N/A |
| └─ExchangeReceiver_24(Probe) | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:273ms, min:0s, p80:273ms, p95:273ms, iters:1618, tasks:3, threads:60} | | N/A | N/A |
| └─ExchangeSender_23 | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:379ms, min:0s, p80:379ms, p95:379ms, iters:568, tasks:3, threads:60} | ExchangeType: HashPartition, Hash Cols: [name: sf10.lineitem.l_linenumber, collate: binary] | N/A | N/A |
| └─TableFullScan_22 | 29999795.00 | 29999795 | mpp[tiflash] | table:b | tiflash_task:{proc max:52ms, min:0s, p80:52ms, p95:52ms, iters:568, tasks:3, threads:60} | keep order:false | N/A | N/A |
+----------------------------------------------------+--------------+-----------+--------------+---------------+--------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------------+----------+------+
16 rows in set (8.07 sec)

mysql> explain analyze select sum(a.L_PARTKEY) from lineitem a left join lineitem b on a.L_ORDERKEY = b.L_LINENUMBER group by a.L_ORDERKEY;
+--------------------------------------------+--------------+-----------+--------------+---------------+--------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------+--------+------+
| id | estRows | actRows | task | access object | execution info | operator info | memory | disk |
+--------------------------------------------+--------------+-----------+--------------+---------------+--------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------+--------+------+
| TableReader_41 | 7623680.00 | 7500000 | root | | time:3.03s, loops:7427, cop_task: {num: 261, max: 0s, min: 0s, avg: 0s, p95: 0s, copr_cache_hit_ratio: 0.00} | data:ExchangeSender_40 | N/A | N/A |
| └─ExchangeSender_40 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:2.88s, min:2.42s, p80:2.88s, p95:2.88s, iters:768, tasks:3, threads:60} | ExchangeType: PassThrough | N/A | N/A |
| └─Projection_39 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:2.74s, min:2.31s, p80:2.74s, p95:2.74s, iters:768, tasks:3, threads:60} | Column#35 | N/A | N/A |
| └─HashAgg_11 | 7623680.00 | 7500000 | mpp[tiflash] | | tiflash_task:{proc max:2.85s, min:2.35s, p80:2.85s, p95:2.85s, iters:768, tasks:3, threads:3} | group by:Column#40, funcs:sum(Column#39)->Column#35 | N/A | N/A |
| └─Projection_52 | 118051610.25 | 137138461 | mpp[tiflash] | | tiflash_task:{proc max:767.4ms, min:638.3ms, p80:767.4ms, p95:767.4ms, iters:2083, tasks:3, threads:60} | cast(sf10.lineitem.l_partkey, decimal(20,0) BINARY)->Column#39, sf10.lineitem.l_orderkey | N/A | N/A |
| └─HashJoin_37 | 118051610.25 | 137138461 | mpp[tiflash] | | tiflash_task:{proc max:823.4ms, min:644.4ms, p80:823.4ms, p95:823.4ms, iters:2083, tasks:3, threads:120} | left outer join, equal:[eq(sf10.lineitem.l_orderkey, sf10.lineitem.l_linenumber)] | N/A | N/A |
| ├─ExchangeReceiver_21(Build) | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:73.3ms, min:62.4ms, p80:73.3ms, p95:73.3ms, iters:1639, tasks:3, threads:60} | | N/A | N/A |
| │ └─ExchangeSender_20 | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:151.8ms, min:0s, p80:151.8ms, p95:151.8ms, iters:576, tasks:3, threads:60} | ExchangeType: HashPartition, Hash Cols: [name: sf10.lineitem.l_orderkey, collate: binary] | N/A | N/A |
| │ └─TableFullScan_19 | 29999795.00 | 29999795 | mpp[tiflash] | table:a | tiflash_task:{proc max:59.8ms, min:0s, p80:59.8ms, p95:59.8ms, iters:576, tasks:3, threads:60} | keep order:false | N/A | N/A |
| └─ExchangeReceiver_24(Probe) | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:215.4ms, min:67.4ms, p80:215.4ms, p95:215.4ms, iters:1603, tasks:3, threads:60} | | N/A | N/A |
| └─ExchangeSender_23 | 29999795.00 | 29999795 | mpp[tiflash] | | tiflash_task:{proc max:467.5ms, min:0s, p80:467.5ms, p95:467.5ms, iters:566, tasks:3, threads:60} | ExchangeType: HashPartition, Hash Cols: [name: sf10.lineitem.l_linenumber, collate: binary] | N/A | N/A |
| └─TableFullScan_22 | 29999795.00 | 29999795 | mpp[tiflash] | table:b | tiflash_task:{proc max:24.5ms, min:0s, p80:24.5ms, p95:24.5ms, iters:566, tasks:3, threads:60} | keep order:false | N/A | N/A |
+--------------------------------------------+--------------+-----------+--------------+---------------+--------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------+--------+------+
12 rows in set (3.04 sec)
```

The difference between the two queries is the first query selects an extra `a.L_ORDERKEY`, an aggregate key.
The result is that a redundant aggregation and exchange are added in execution plan, which makes the first query two times slower than the second one(6s vs 3s in debug version).

### 2. What did you expect to see? (Required)
No redundant aggregation and exchange.

### 3. What did you see instead (Required)
Redundant aggregation and exchange.

### 4. What is your TiDB version? (Required)
v6.1.0