add pclmulqdq crc32 variant

[folkertdev]

We've had some discussion about whether we really need to have our own implementation of crc32. It turns out that we get a massive speed boost over crc32fast in the streaming case (when input comes in in small chunks):

Benchmark 1 (84 runs): cargo run --release --example crc32_bench sse-chunked silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          59.8ms ± 3.60ms    57.6ms … 84.2ms          6 ( 7%)        0%
  peak_rss           21.9MB ±  171KB    21.5MB … 22.3MB          0 ( 0%)        0%
  cpu_cycles          159M  ± 7.31M      155M  …  206M           6 ( 7%)        0%
  instructions        351M  ± 37.4K      351M  …  351M           0 ( 0%)        0%
  cache_references   3.76M  ± 63.5K     3.65M  … 3.97M           3 ( 4%)        0%
  cache_misses       1.06M  ± 81.3K      908K  … 1.49M           8 (10%)        0%
  branch_misses       750K  ± 59.1K      730K  … 1.17M           4 ( 5%)        0%
Benchmark 2 (56 runs): cargo run --release --example crc32_bench crc32fast-chunked silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          90.8ms ± 2.10ms    88.4ms …  101ms          2 ( 4%)        💩+ 51.7% ±  1.7%
  peak_rss           21.9MB ±  170KB    21.5MB … 22.2MB          0 ( 0%)          -  0.1% ±  0.3%
  cpu_cycles          278M  ± 4.38M      273M  …  302M           2 ( 4%)        💩+ 75.4% ±  1.3%
  instructions        434M  ± 35.6K      434M  …  434M           0 ( 0%)        💩+ 23.5% ±  0.0%
  cache_references   3.75M  ± 54.5K     3.63M  … 3.90M           0 ( 0%)          -  0.1% ±  0.5%
  cache_misses       1.14M  ±  121K      992K  … 1.58M           3 ( 5%)        💩+  7.7% ±  3.2%
  branch_misses       744K  ± 15.3K      732K  …  842K           3 ( 5%)          -  0.9% ±  2.1%

when all the input is availabe, the current implementation is just as fast as crc32fast.

[codecov]

Codecov Report

Attention: 63 lines in your changes are missing coverage. Please review.

Comparison is base (92994f8) 85.99% compared to head (0297489) 86.16%.
Report is 2 commits behind head on main.

Files	Patch %	Lines
zlib-rs/src/crc32/braid.rs	82.30%	20 Missing ⚠️
zlib-rs/src/crc32/pclmulqdq.rs	92.88%	19 Missing ⚠️
zlib-rs/src/crc32.rs	84.84%	15 Missing ⚠️
load-dynamic-libz-ng/src/lib.rs	0.00%	9 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #31      +/-   ##
==========================================
+ Coverage   85.99%   86.16%   +0.16%     
==========================================
  Files          29       31       +2     
  Lines        6591     6982     +391     
==========================================
+ Hits         5668     6016     +348     
- Misses        923      966      +43     

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

[bramtweedegolf]

Maybe add this to the repo?

[folkertdev]

will do in a separate PR