Remove redundant exchange if its hash key set contains the previous exchange's hash key set

## Enhancement

TPC-H 100G query 3
```sql
+--------------------------------------------------------------+--------------+--------------+----------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| id | estRows | task | access object | operator info |
+--------------------------------------------------------------+--------------+--------------+----------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Projection_14 | 10.00 | root | | tpch_100_multi_key.lineitem.l_orderkey, Column#35, tpch_100_multi_key.orders.o_orderdate, tpch_100_multi_key.orders.o_shippriority |
| └─TopN_18 | 10.00 | root | | Column#35:desc, tpch_100_multi_key.orders.o_orderdate, offset:0, count:10 |
| └─TableReader_153 | 10.00 | root | | data:ExchangeSender_152 |
| └─ExchangeSender_152 | 10.00 | mpp[tiflash] | | ExchangeType: PassThrough |
| └─TopN_151 | 10.00 | mpp[tiflash] | | Column#35:desc, tpch_100_multi_key.orders.o_orderdate, offset:0, count:10 |
| └─Projection_146 | 81356609.94 | mpp[tiflash] | | Column#35, tpch_100_multi_key.orders.o_orderdate, tpch_100_multi_key.orders.o_shippriority, tpch_100_multi_key.lineitem.l_orderkey |
| └─HashAgg_144 | 81356609.94 | mpp[tiflash] | | group by:Column#64, Column#65, Column#66, funcs:sum(Column#60)->Column#35, funcs:firstrow(Column#61)->tpch_100_multi_key.orders.o_orderdate, funcs:firstrow(Column#62)->tpch_100_multi_key.orders.o_shippriority, funcs:firstrow(Column#63)->tpch_100_multi_key.lineitem.l_orderkey |
| └─Projection_154 | 182757407.58 | mpp[tiflash] | | mul(tpch_100_multi_key.lineitem.l_extendedprice, minus(1, tpch_100_multi_key.lineitem.l_discount))->Column#60, tpch_100_multi_key.orders.o_orderdate, tpch_100_multi_key.orders.o_shippriority, tpch_100_multi_key.lineitem.l_orderkey, tpch_100_multi_key.lineitem.l_orderkey, tpch_100_multi_key.orders.o_orderdate, tpch_100_multi_key.orders.o_shippriority |
| └─ExchangeReceiver_121 | 182757407.58 | mpp[tiflash] | | |
| └─ExchangeSender_120 | 182757407.58 | mpp[tiflash] | | ExchangeType: HashPartition, Hash Cols: [name: tpch_100_multi_key.lineitem.l_orderkey, collate: binary], [name: tpch_100_multi_key.orders.o_orderdate, collate: binary], [name: tpch_100_multi_key.orders.o_shippriority, collate: binary] |
| └─HashJoin_119 | 182757407.58 | mpp[tiflash] | | inner join, equal:[eq(tpch_100_multi_key.orders.o_orderkey, tpch_100_multi_key.lineitem.l_orderkey)] |
| ├─ExchangeReceiver_59(Build) | 45954571.12 | mpp[tiflash] | | |
| │ └─ExchangeSender_58 | 45954571.12 | mpp[tiflash] | | ExchangeType: HashPartition, Hash Cols: [name: tpch_100_multi_key.orders.o_orderkey, collate: binary] |
| │ └─HashJoin_50 | 45954571.12 | mpp[tiflash] | | inner join, equal:[eq(tpch_100_multi_key.customer.c_custkey, tpch_100_multi_key.orders.o_custkey)] |
| │ ├─ExchangeReceiver_54(Build) | 3023286.28 | mpp[tiflash] | | |
| │ │ └─ExchangeSender_53 | 3023286.28 | mpp[tiflash] | | ExchangeType: Broadcast |
| │ │ └─Selection_52 | 3023286.28 | mpp[tiflash] | | eq(tpch_100_multi_key.customer.c_mktsegment, "AUTOMOBILE") |
| │ │ └─TableFullScan_51 | 15000000.00 | mpp[tiflash] | table:customer | keep order:false |
| │ └─Selection_57(Probe) | 72631642.00 | mpp[tiflash] | | lt(tpch_100_multi_key.orders.o_orderdate, 1995-03-13 00:00:00.000000) |
| │ └─TableFullScan_56 | 150000000.00 | mpp[tiflash] | table:orders | keep order:false |
| └─ExchangeReceiver_64(Probe) | 323548294.78 | mpp[tiflash] | | | |
| └─ExchangeSender_63 | 323548294.78 | mpp[tiflash] | | ExchangeType: HashPartition, Hash Cols: [name: tpch_100_multi_key.lineitem.l_orderkey, collate: binary] |
| └─Selection_62 | 323548294.78 | mpp[tiflash] | | gt(tpch_100_multi_key.lineitem.l_shipdate, 1995-03-13 00:00:00.000000) |
| └─TableFullScan_61 | 600037902.00 | mpp[tiflash] | table:lineitem | keep order:false |
+--------------------------------------------------------------+--------------+--------------+----------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
```
After the `HashJoin_119`, the data has already partitioned by `l_orderkey`(or `o_orderkey`) in different TiFlash nodes.
Then `ExchangeSender_120` does a hash partition by `l_orderkey`, `o_orderdate` and `o_shippriority`, which is used for aggregate function.
In fact, `ExchangeSender_120` can be removed and the `HashAgg` can compute locally.
It's because all records with the same `l_orderkey`, `o_orderdate` and `o_shippriority` must be located on the same TiFlash node after `HashJoin_119`.

So in general, if the hash key set contains the previous exchange's hash key set, the exchange can be removed.
Note that this rule can also be used for hash join.