Filter push down for Union

[Dandandan]

Which issue does this PR close?

Closes #557

Rationale for this change

What changes are included in this PR?

Filter is pushed down through union (all), so it can be pushed down further towards other operations like table scans.

Are there any user-facing changes?

FYI @nevi-me

[codecov-commenter]

Codecov Report

Merging #559 (1385342) into master (d382854) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #559   +/-   ##
=======================================
  Coverage   76.09%   76.10%           
=======================================
  Files         156      156           
  Lines       27047    27056    +9     
=======================================
+ Hits        20581    20590    +9     
  Misses       6466     6466           

Impacted Files	Coverage Δ
datafusion/src/optimizer/filter_push_down.rs	97.84% <100.00%> (+0.04%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d382854...1385342. Read the comment docs.

[nevi-me]

I'm seeing:

Projection: #COUNT(UInt8(1)) AS total_records, #COUNT(DISTINCT payment_type) AS total_payment_types, #SUM(CAST(trip_distance AS Float64)) AS total_distance
  Aggregate: groupBy=[[]], aggr=[[COUNT(UInt8(1)), COUNT(DISTINCT #payment_type), SUM(CAST(#trip_distance AS Float64))]]
    Union
      Projection: #passenger_count, #trip_distance, #payment_type, #total_amount
        TableScan: mongo_nyc projection=Some([3, 4, 9, 16]), filters=[#passenger_count Gt Int64(3), #total_amount Lt Float64(20)]
      Projection: #passenger_count, #trip_distance, #payment_type, #total_amount
        Filter: #passenger_count Gt Int64(3) And #total_amount Lt Float64(20)
          TableScan: csv_nyc projection=Some([3, 4, 9, 16])

Thanks @Dandandan