folly::crc32c faster than absl::crc32c for some sizes

[MBkkt]

cpu: AMD Ryzen 9 7950X 16-Core Processor
clang: 20
libc++: 19
options: O3, avx2 (haswell like, but avx512 available for avx512_crc32c target and checked in runtime for folly sizes >= 4097)
abseil and folly versions are close to current master

For some sizes, approximately (129..255] and [2048..4096], folly function is a little faster.
For small sizes abseil is significantly faster (because have better branches for small sizes)
For large sizes abseil is also significantly faster (don't know why, probably because CPU specific count of pcmul streams)

folly implementation is changed from 4097 (included), (avx2 folly slower than avx512)

Run on (32 X 5010.09 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x16)
  L1 Instruction 32 KiB (x16)
  L2 Unified 1024 KiB (x16)
  L3 Unified 32768 KiB (x2)
Load Average: 1.13, 0.36, 0.87
-------------------------------------------------------------------
Benchmark                         Time             CPU   Iterations
-------------------------------------------------------------------
BM_CalculateAbseil/0          0.646 ns        0.646 ns   1088247328
BM_CalculateAbseil/1          0.756 ns        0.756 ns    929241117
BM_CalculateAbseil/2          0.744 ns        0.744 ns    941916174
BM_CalculateAbseil/3          0.865 ns        0.865 ns    798610637
BM_CalculateAbseil/4          0.645 ns        0.645 ns   1086374687
BM_CalculateAbseil/5          0.917 ns        0.917 ns    763600758
BM_CalculateAbseil/7          0.980 ns        0.980 ns    715095940
BM_CalculateAbseil/8          0.731 ns        0.731 ns    957964504
BM_CalculateAbseil/9          0.919 ns        0.919 ns    762666864
BM_CalculateAbseil/15          1.11 ns         1.11 ns    626702718
BM_CalculateAbseil/16         0.913 ns        0.913 ns    766585199
BM_CalculateAbseil/17          1.10 ns         1.10 ns    635899395
BM_CalculateAbseil/31          1.53 ns         1.53 ns    458903386
BM_CalculateAbseil/32          1.28 ns         1.28 ns    547584695
BM_CalculateAbseil/33          1.47 ns         1.47 ns    476916583
BM_CalculateAbseil/63          2.51 ns         2.51 ns    279598900
BM_CalculateAbseil/64          2.00 ns         2.00 ns    350106994
BM_CalculateAbseil/65          2.63 ns         2.63 ns    266499836
BM_CalculateAbseil/127         5.40 ns         5.40 ns    130181667
BM_CalculateAbseil/128         4.52 ns         4.52 ns    155017426
BM_CalculateAbseil/129         4.70 ns         4.70 ns    149218064
BM_CalculateAbseil/255         12.3 ns         12.2 ns     57085432
BM_CalculateAbseil/256         8.44 ns         8.44 ns     82957778
BM_CalculateAbseil/257         8.53 ns         8.53 ns     82192356
BM_CalculateAbseil/511         14.1 ns         14.1 ns     49496599
BM_CalculateAbseil/512         13.8 ns         13.8 ns     50872816
BM_CalculateAbseil/513         14.0 ns         14.0 ns     50200548
BM_CalculateAbseil/1023        26.5 ns         26.5 ns     26781396
BM_CalculateAbseil/1024        25.8 ns         25.8 ns     27180258
BM_CalculateAbseil/1025        26.6 ns         26.6 ns     26650233
BM_CalculateAbseil/2047        49.8 ns         49.8 ns     13878418
BM_CalculateAbseil/2048        55.4 ns         55.4 ns     12622499
BM_CalculateAbseil/2049        61.7 ns         61.7 ns     11328466
BM_CalculateAbseil/4095         109 ns          109 ns      6395921
BM_CalculateAbseil/4096        99.5 ns         99.5 ns      7039485
BM_CalculateAbseil/4097         105 ns          105 ns      6688720
BM_CalculateAbseil/8191         204 ns          204 ns      3431597
BM_CalculateAbseil/8192         195 ns          195 ns      3582733
BM_CalculateAbseil/8193         202 ns          202 ns      3469483
BM_CalculateAbseil/16383        390 ns          390 ns      1793741
BM_CalculateAbseil/16384        380 ns          380 ns      1838427
BM_CalculateAbseil/16385        387 ns          387 ns      1810665
BM_CalculateFolly/0            1.85 ns         1.85 ns    377662494
BM_CalculateFolly/1            2.04 ns         2.04 ns    343704313
BM_CalculateFolly/2            2.22 ns         2.22 ns    315298626
BM_CalculateFolly/3            2.04 ns         2.04 ns    342723437
BM_CalculateFolly/4            2.40 ns         2.40 ns    292177743
BM_CalculateFolly/5            2.40 ns         2.40 ns    291749153
BM_CalculateFolly/7            2.05 ns         2.05 ns    342217301
BM_CalculateFolly/8            2.41 ns         2.41 ns    290835577
BM_CalculateFolly/9            2.59 ns         2.59 ns    270002300
BM_CalculateFolly/15           2.63 ns         2.63 ns    267572419
BM_CalculateFolly/16           3.22 ns         3.22 ns    209087870
BM_CalculateFolly/17           3.02 ns         3.02 ns    198077036
BM_CalculateFolly/31           3.26 ns         3.26 ns    211032022
BM_CalculateFolly/32           3.32 ns         3.32 ns    211045834
BM_CalculateFolly/33           3.32 ns         3.32 ns    210885730
BM_CalculateFolly/63           3.52 ns         3.52 ns    198911325
BM_CalculateFolly/64           3.15 ns         3.15 ns    222390235
BM_CalculateFolly/65           3.33 ns         3.33 ns    210027665
BM_CalculateFolly/127          5.75 ns         5.75 ns    121163035
BM_CalculateFolly/128          5.11 ns         5.11 ns    136833222
BM_CalculateFolly/129          5.38 ns         5.38 ns    130043580
BM_CalculateFolly/255          9.21 ns         9.21 ns     75215046
BM_CalculateFolly/256          8.95 ns         8.95 ns     78224445
BM_CalculateFolly/257          8.99 ns         8.99 ns     78205576
BM_CalculateFolly/511          15.0 ns         15.0 ns     46608749
BM_CalculateFolly/512          14.9 ns         14.9 ns     47155267
BM_CalculateFolly/513          15.0 ns         15.0 ns     46609166
BM_CalculateFolly/1023         26.3 ns         26.3 ns     26564993
BM_CalculateFolly/1024         26.2 ns         26.2 ns     26762909
BM_CalculateFolly/1025         26.1 ns         26.1 ns     26781352
BM_CalculateFolly/2047         49.2 ns         49.2 ns     14238226
BM_CalculateFolly/2048         49.0 ns         49.0 ns     14245964
BM_CalculateFolly/2049         49.2 ns         49.2 ns     14209606
BM_CalculateFolly/4095         96.7 ns         96.7 ns      7245278
BM_CalculateFolly/4096         96.3 ns         96.3 ns      7270017
BM_CalculateFolly/4097          143 ns          143 ns      4891778
BM_CalculateFolly/8191          269 ns          269 ns      2607869
BM_CalculateFolly/8192          261 ns          261 ns      2680527
BM_CalculateFolly/8193          262 ns          262 ns      2666659
BM_CalculateFolly/16383         496 ns          496 ns      1410627
BM_CalculateFolly/16384         493 ns          493 ns      1419063
BM_CalculateFolly/16385         494 ns          494 ns      1418124

Such patch fix it for this CPU, maybe even 4097 (aka 4096 + >= instead of >) better

In general diff isn't large, so it's more just curiosity how these values was chosen and why didn't you have more sizes around these branch points in crc32c microbenchmark.

Also abseil already have CPU specific implementation so maybe it make sense to do:

1. these sizes also CPU specific
2. add more sizes in benchmark
3. find best values for each CPU