[EXP] use roaring bitmaps for MinHash

[luizirber]

Try out using roaring bitmaps as in-memory representation of hashes for MinHash. We already use roaring bitmaps in the RevIndex (to store lists of datasets containing a specific hash).

This PR mostly goes for minimal changes to make it work, but future goals are

• optimize, this is just enough to make it work
• serialize/deserialize directly from disk, without generating a full list of hashes (this will likely need to move away from JSON...)
• try out for LargeMinHash too? see how bad it is

[codspeed-hq]

CodSpeed Performance Report

Merging #3444 will degrade performances by 94.34%

_{Comparing lirber/roaring_mh (f07cba6) with latest (f4f5187)}

Summary

⚡ 1 improvements
❌ 7 regressions
✅ 13 untouched benchmarks

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

Benchmarks breakdown

	Benchmark	latest	lirber/roaring_mh	Change
❌	abund0_ani_ci0	717 µs	12,671.4 µs	-94.34%
❌	abund0_ani_ci1	736.7 µs	12,834.6 µs	-94.26%
❌	abund1_ani_ci0	938.6 µs	14,355.9 µs	-93.46%
❌	abund1_ani_ci1	960 µs	14,518.1 µs	-93.39%
❌	intersection	20.3 µs	200.1 µs	-89.88%
❌	intersection_size	18.3 µs	157.2 µs	-88.36%
❌	large intersection	10.6 ms	22.7 ms	-53.27%
⚡	large intersection_size	9,302.8 µs	234.2 µs	×40

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.38%. Comparing base (f4f5187) to head (f07cba6).

Additional details and impacted files

@@            Coverage Diff             @@
##           latest    #3444      +/-   ##
==========================================
+ Coverage   86.42%   92.38%   +5.95%     
==========================================
  Files         137      104      -33     
  Lines       16106    12925    -3181     
  Branches     2219     2219              
==========================================
- Hits        13920    11941    -1979     
+ Misses       1879      676    -1203     
- Partials      307      308       +1     

Flag	Coverage Δ
hypothesis-py	25.43% <ø> (ø)
python	92.38% <ø> (-0.02%)	⬇️
rust	?

Flags with carried forward coverage won't be shown. Click here to find out more.

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