Wrong result when aggregating on volatile functions

## Bug Report

```sql
create table foo(a int, b int, c int, d int);
insert into foo values(1,1,1,10),(2,2,2,2);
insert into foo values(3,1,1,10),(3,2,2,2);

explain select b*floor(2*rand()) as e, count(d) from foo group by e;
+-----------------------------+---------+--------------+---------------+------------------------------------------------------------------------------------------------+
| id | estRows | task | access object | operator info |
+-----------------------------+---------+--------------+---------------+------------------------------------------------------------------------------------------------+
| Projection_4 | 3.20 | root | | mul(cast(test.foo.b, double BINARY), floor(mul(2, rand())))->Column#7, Column#6 |
| └─HashAgg_5 | 3.20 | root | | group by:Column#11, funcs:count(Column#9)->Column#6, funcs:firstrow(Column#10)->test.foo.b |
| └─Projection_15 | 4.00 | root | | test.foo.d, test.foo.b, mul(cast(test.foo.b, double BINARY), floor(mul(2, rand())))->Column#11 |
| └─TableReader_10 | 4.00 | root | | data:TableFullScan_9 |
| └─TableFullScan_9 | 4.00 | cop[tiflash] | table:foo | keep order:false, stats:pseudo |
+-----------------------------+---------+--------------+---------------+------------------------------------------------------------------------------------------------+
5 rows in set, 4 warnings (0.00 sec)

-- sometimes I see this kind of suspicious result
select b*floor(2*rand()) as e, count(d) from foo group by e;
+------+----------+
| e | count(d) |
+------+----------+
| 2 | 1 |
| 2 | 1 |
| 0 | 2 |
+------+----------+
3 rows in set (0.02 sec)
```
Shouldn't each aggregate key only appear once?

I think the culprit may be we do the project again after the HashAgg. Each time rand() will generate different values.