feat: Implement single inequality joins for join_where

[adamreeve]

This implements the "piecewise merge join" algorithm (described in this DuckDB article) to handle join_where with a single inequality, without using a cross join and filter.

I've reused the IEJoin join type internally due to the similarities in how these two join types are handled. Technically this isn't using the IEJoin algorithm but it is another type of inequality join, so this seems OK, but it could be pulled out into its own join type if needed.

[codspeed-hq]

CodSpeed Performance Report

Merging #18727 will degrade performances by 25.3%

_{Comparing adamreeve:iejoin_single (f45d106) with main (54218e7)}

Summary

❌ 1 regressions
✅ 38 untouched benchmarks

🆕 1 new benchmarks

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

Benchmarks breakdown

	Benchmark	main	adamreeve:iejoin_single	Change
❌	test_groupby_h2oai_q3	2.3 ms	3.1 ms	-25.3%
🆕	test_single_inequality	N/A	80.8 ms	N/A

[codecov]

Codecov Report

Attention: Patch coverage is 86.95652% with 24 lines in your changes missing coverage. Please review.

Project coverage is 79.86%. Comparing base (dddf0b7) to head (f45d106).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-ops/src/frame/join/iejoin/mod.rs	85.36%	24 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #18727      +/-   ##
==========================================
- Coverage   79.88%   79.86%   -0.02%     
==========================================
  Files        1513     1513              
  Lines      203466   203631     +165     
  Branches     2892     2892              
==========================================
+ Hits       162546   162640      +94     
- Misses      40372    40443      +71     
  Partials      548      548              

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

[ritchie46]

Nice one @adamreeve. Great that we can reuse the parallelism logic. Makes a lot of sense.

All this work can also greatly be re-used in the streaming engine.