Follow-up from #12908

Allow dynamically generated join filters to be pushed through UNION, UNNEST and AGGREGATE nodes in the plan.

For UNION, this means we need to potentially push filters into multiple table scans. For example, in this case we push the filter into both scans over lineitem:

CREATE VIEW double_lineitem AS
FROM lineitem
UNION ALL
FROM lineitem;

SELECT *
FROM double_lineitem
WHERE l_orderkey=(SELECT MAX(l_orderkey) FROM lineitem) ORDER BY ALL

We can push filters through aggregates as well, as long as the filters are on grouping columns. However, in this case, the pipeline dependencies are not correctly set up yet to ensure the join upstream is executed before the aggregate - meaning the filter has no effect yet as the aggregate is computed before we know which rows we can filter out (CC @lnkuiper perhaps something you want to look into at some point).