Hi, hackers
I propose to slightly improve the performance of nested loop join in the case of partitioned inner table.
As I see in the code, the backend looks for the partition of the inner table each time after fetch a new row from the outer table.
These searches can take a significant amount of time.
But we can skip this step if the nested loop parameter(s) was(re) not changed since the previous row fetched from the outer table
The draft patch is attached.
