wasm: SIMD slide_hash

[folkertdev]

here we are comparing miniz_oxide at level 6 with zlib-rs at level 7. At those settings, they produce output of roughly equal size.

> wasmtime target/wasm32-wasi/release/wasm-zlib-benchmark.wasm deflate miniz_oxide
[src/main.rs:46:17] vec.len() = 6411455
> wasmtime target/wasm32-wasi/release/wasm-zlib-benchmark.wasm deflate zlib-rs
[src/main.rs:57:17] out.len() = 6416929

For these settings, our baseline is

Benchmark 1 (8 runs): wasmtime target/wasm32-wasi/release/wasm-zlib-benchmark.wasm deflate miniz_oxide
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           694ms ± 46.3ms     677ms …  809ms          1 (13%)        0%
  peak_rss           88.1MB ± 16.8MB    82.1MB …  130MB          2 (25%)        0%
  cpu_cycles         3.07G  ±  496M     2.89G  … 4.30G           1 (13%)        0%
  instructions       7.05G  ±  620M     6.83G  … 8.59G           1 (13%)        0%
  cache_references    304M  ± 29.2M      293M  …  376M           1 (13%)        0%
  cache_misses       4.63M  ± 7.78M     1.67M  … 23.9M           1 (13%)        0%
  branch_misses      19.4M  ± 3.18M     18.2M  … 27.2M           2 (25%)        0%
Benchmark 2 (11 runs): wasmtime target/wasm32-wasi/release/wasm-zlib-benchmark.wasm deflate zlib-rs
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           489ms ± 2.45ms     486ms …  494ms          0 ( 0%)        ⚡- 29.6% ±  4.2%
  peak_rss           82.2MB ± 1.23KB    82.1MB … 82.2MB          1 ( 9%)          -  6.7% ± 12.0%
  cpu_cycles         2.00G  ± 11.6M     1.99G  … 2.03G           0 ( 0%)        ⚡- 34.8% ± 10.2%
  instructions       5.56G  ± 1.85K     5.56G  … 5.56G           0 ( 0%)        ⚡- 21.2% ±  5.5%
  cache_references    171M  ± 1.27M      169M  …  173M           0 ( 0%)        ⚡- 43.9% ±  6.1%
  cache_misses       3.75M  ±  982K     2.29M  … 5.19M           0 ( 0%)          - 19.1% ± 106.9%
  branch_misses      12.3M  ± 4.03K     12.3M  … 12.3M           0 ( 0%)        ⚡- 36.5% ± 10.3%

The changes in this PR are performance-neutral for our current inputs