[Feature]: task skew in nested loop join

[zhangyuqin1998]

Use Case

When the data volume is small, only a few granules will be partitioned, which leads to some threads being idle. In the nested loop join scenario, load skew will seriously slow down the execution speed.
Describe the solution you'd like

Inserting a shuffle operator above the left table in a nested loop join balances the load.
Describe alternatives you've considered

Additional context

[hnwyllmm]

Assign it to you.