Planner generated plan that shuffles on single agg key

## Bug Report

```sql
create table foo(a int, b int, c int, d int);
alter table foo set tiflash replica 2;
explain select b*10 as e, c, count(d) from foo group by e,c;
+----------------------------------------+----------+-------------------+---------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------+
| id | estRows | task | access object | operator info |
+----------------------------------------+----------+-------------------+---------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Projection_4 | 8000.00 | root | | mul(tpch_100.foo.b, 10)->Column#7, tpch_100.foo.c, Column#6 |
| └─TableReader_31 | 8000.00 | root | | data:ExchangeSender_30 |
| └─ExchangeSender_30 | 8000.00 | batchCop[tiflash] | | ExchangeType: PassThrough |
| └─Projection_26 | 8000.00 | batchCop[tiflash] | | Column#6, tpch_100.foo.b, tpch_100.foo.c |
| └─HashAgg_27 | 8000.00 | batchCop[tiflash] | | group by:Column#13, tpch_100.foo.c, funcs:sum(Column#14)->Column#6, funcs:firstrow(Column#15)->tpch_100.foo.b, funcs:firstrow(tpch_100.foo.c)->tpch_100.foo.c |
| └─ExchangeReceiver_29 | 8000.00 | batchCop[tiflash] | | |
| └─ExchangeSender_28 | 8000.00 | batchCop[tiflash] | | ExchangeType: HashPartition, Hash Cols: [name: tpch_100.foo.c, collate: binary] |
| └─HashAgg_9 | 8000.00 | batchCop[tiflash] | | group by:Column#23, Column#24, funcs:count(Column#21)->Column#14, funcs:firstrow(Column#22)->Column#15 |
| └─Projection_35 | 10000.00 | batchCop[tiflash] | | tpch_100.foo.d, tpch_100.foo.b, mul(tpch_100.foo.b, 10)->Column#23, tpch_100.foo.c |
| └─TableFullScan_25 | 10000.00 | batchCop[tiflash] | table:foo | keep order:false, stats:pseudo |
+----------------------------------------+----------+-------------------+---------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------+
10 rows in set (0.16 sec)
```

The exchange `ExchangeType: HashPartition, Hash Cols: [name: tpch_100.foo.c, collate: binary]`. The hash cols should be 2 columns instead of a single column `c`.

It is not wrong plan, but an inefficient plan, which may generate data skew.