Check FILTER dedup and FILTER/CASE IR equivalence from #467

[milindsrivastava1997]

Follow-up to #467 (per-measure FILTER predicates on Aggregate, #466). These review questions were not answered before merge: #467 (review)

Questions

1. When several aggregates share the same FILTER predicate, what happens? Do we deduplicate the filters?
2. Do these equivalent queries lower to the same IR?
   • one aggregate with FILTER (WHERE p), e.g. SUM(x) FILTER (WHERE p)
   • one aggregate with the matching CASE, e.g. SUM(CASE WHEN p THEN x END)

Expected outcome

• Write down the current behavior for both questions, with the IR for each case.
• If the FILTER and CASE forms lower to different IR, add a canonicalization pass that gives them the same IR, reusing the canonicalization logic already in the codebase. Add a test that the two forms are equivalent.