Note: migrated from original JIRA: https://issues.apache.org/jira/browse/ARROW-12266

Improve null handling of 

SELECT id1, id2 FROM (SELECT null AS id1) t1
INNER JOIN (SELECT 0 AS id2) t2 ON id1 = id2

NULL, NULL

(should be empty result set)

We should filter beforehand to make this result correct. Also this can make things more efficient as the non-null filter can be pushed down which can lead to efficiency gains (making data-set smaller, not having to deal with nullable data, or even entire files could be skipped when they only contain nulls).