Performance issue on `group by having min(xxx)`

[actiontech-bot]

Raised by: helingyun Steps:

1. Prepare data: 2 nodes, 2 tables on each node, 50000000 rows in each table
2. Sysbench 2048-thread queries: select min(my_max_money),my_creditcard,my_comsume_money from user_creditcard_info where id between 1 and 100 group by my_creditcard having min(my_comsume_money)>=8 ;