That is, if table `foo` and table `bar` are both tables on the same
remote server, then when I do `select * from foo, bar`, can it
delegate the whole query on the remote side, rather than fetching rows
from both servers one by one and do merging on the local side?
For example:
```
foo=> explain select * from foreign_test2, foreign_test where m = id;
QUERY PLAN
-------------------------------------------------------------------------------------
Merge Join (cost=444.06..590.63 rows=9316 width=72)
Merge Cond: (foreign_test2.m = foreign_test.id)
-> Sort (cost=222.03..225.44 rows=1365 width=36)
Sort Key: foreign_test2.m
-> Foreign Scan on foreign_test2 (cost=100.00..150.95
rows=1365 width=36)
-> Sort (cost=222.03..225.44 rows=1365 width=36)
Sort Key: foreign_test.id
-> Foreign Scan on foreign_test (cost=100.00..150.95
rows=1365 width=36)
```
If the planning involving remote tables, could the planner push down
the queries once to gain the best efficiency?
