Improve intersect_by_rank performance

[robert3005]

Intersect by rank is an operation for performing filter on top of another filter

This pr improves
Dense on Dense masks: 10-18x
Sparse on Dense masks: 8x
Sparse on Sparse masks: 2x

Signed-off-by: Robert Kruszewski github@robertk.io

[robert3005]

I will go over this tomorrow @joseph-isaacs I looked at #7098 and #7393 which both optimised slightly different cases of this function. I tried to combine the two.

[codspeed-hq]

Merging this PR will degrade performance by 23.81%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 24 improved benchmarks
❌ 24 regressed benchmarks
✅ 1160 untouched benchmarks

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

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	take_search[(0.005, 0.05)]	132 µs	168.3 µs	-21.6%
❌	Simulation	take_search[(0.005, 0.1)]	247.5 µs	320.3 µs	-22.72%
❌	Simulation	take_search[(0.005, 0.5)]	1.2 ms	1.5 ms	-23.67%
❌	Simulation	take_search[(0.005, 1.0)]	2.3 ms	3.1 ms	-23.81%
❌	Simulation	take_search[(0.01, 0.05)]	143.1 µs	179.4 µs	-20.27%
❌	Simulation	take_search[(0.01, 0.1)]	268.4 µs	341.2 µs	-21.33%
❌	Simulation	take_search[(0.01, 0.5)]	1.3 ms	1.6 ms	-22.21%
❌	Simulation	take_search[(0.01, 1.0)]	2.5 ms	3.3 ms	-22.34%
❌	Simulation	take_search[(0.1, 0.05)]	212.7 µs	249.1 µs	-14.6%
❌	Simulation	take_search[(0.1, 0.1)]	385.9 µs	458.6 µs	-15.87%
❌	Simulation	take_search[(0.1, 0.5)]	1.8 ms	2.2 ms	-16.93%
❌	Simulation	take_search[(0.1, 1.0)]	3.5 ms	4.3 ms	-17.09%
❌	Simulation	take_search_chunked[(0.005, 0.05)]	162.2 µs	193.2 µs	-16.01%
❌	Simulation	take_search_chunked[(0.005, 0.1)]	307.3 µs	369.2 µs	-16.76%
❌	Simulation	take_search_chunked[(0.005, 0.5)]	1.5 ms	1.8 ms	-17.39%
❌	Simulation	take_search_chunked[(0.005, 1.0)]	2.9 ms	3.5 ms	-17.48%
❌	Simulation	take_search_chunked[(0.01, 0.05)]	175.3 µs	206.2 µs	-14.99%
❌	Simulation	take_search_chunked[(0.01, 0.1)]	332.1 µs	393.9 µs	-15.7%
❌	Simulation	take_search_chunked[(0.01, 0.5)]	1.6 ms	1.9 ms	-16.3%
❌	Simulation	take_search_chunked[(0.01, 1.0)]	3.2 ms	3.8 ms	-16.38%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---

_{Comparing rk/intersect-by-rank (8578962) with develop (823991f)}

[joseph-isaacs]

add a todo to add a sized iterator variant, then you can switch over the ratio of len to indices

[joseph-isaacs]

Add benchmark numbers to the desc