Use .as_chunks_mut() instead of .chunks_exact_mut() for better performance

[Shnatsel]

Gives a significant boost on the benchmarks in the mid-range sizes - big enough to use the large variable size codepath, but small enough not to be entirely bottlenecked on memory access.

Supersedes #15

Benchmarks on desktop Zen4: https://gist.github.com/Shnatsel/1d700a953ba94435bfa53415fc671ed4

Benchmarked using RUSTFLAGS='-C target-cpu=native' cargo bench 'PhastFT' --bench=bench -- --baseline=main-native

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.29%. Comparing base (7dc39b1) to head (6f6ff7d).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #51   +/-   ##
=======================================
  Coverage   99.29%   99.29%           
=======================================
  Files          12       12           
  Lines        2257     2277   +20     
=======================================
+ Hits         2241     2261   +20     
  Misses         16       16           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Shnatsel]

Since this speeds up the part that crosses both halves, this makes the gains from #50 even greater, with −73.529% time for Forward f64/PhastFT DIT/16777216 or close to 4x speedup