perf: Remove predicate from IR::DataFrame

[ritchie46]

Simplifies IR, saves a lowering step in streaming and ensures that the in-memory engine can parallelize if it wants to.

First solution for #20475

[codspeed-hq]

CodSpeed Performance Report

Merging #20492 will degrade performances by 55.9%

_{Comparing filter (b743532) with main (f43a7d4)}

Summary

❌ 8 regressions
✅ 33 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	filter	Change
❌	test_pdsh_q12	7.5 ms	10.2 ms	-25.72%
❌	test_pdsh_q13	10.6 ms	24 ms	-55.9%
❌	test_pdsh_q14	3.4 ms	4.7 ms	-26.61%
❌	test_pdsh_q15	3.1 ms	4.5 ms	-30.24%
❌	test_pdsh_q16	7.8 ms	14.7 ms	-46.66%
❌	test_pdsh_q3	7.5 ms	10.6 ms	-29.53%
❌	test_pdsh_q4	5.9 ms	8.1 ms	-27.47%
❌	test_pdsh_q6	3.7 ms	5.5 ms	-32.65%

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 79.01%. Comparing base (f43a7d4) to head (b743532).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #20492      +/-   ##
==========================================
- Coverage   79.02%   79.01%   -0.01%     
==========================================
  Files        1563     1563              
  Lines      220596   220532      -64     
  Branches     2492     2492              
==========================================
- Hits       174317   174249      -68     
- Misses      45706    45710       +4     
  Partials      573      573              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.