perf: reduce register spillage by applying twiddle symmetry

[smu160]

No description provided.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.80%. Comparing base (3cebc5d) to head (fdb57be).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #112      +/-   ##
==========================================
- Coverage   98.87%   98.80%   -0.08%     
==========================================
  Files          10       10              
  Lines        2227     2094     -133     
==========================================
- Hits         2202     2069     -133     
  Misses         25       25              

☔ 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]

I think there are really two changes here: using narrower vectors (which makes sense for radix-2^2) and using the symmetry.

We should revisit this older codelet version when we add AVX-512, where we can use 512-bit vectors natively for math and also not worry about register pressure due to 2x wider registers and 2x the number of registers.

[Shnatsel]

I'm seeing better results from #113 which supersedes this. So I'll likely clean up #113 and merge that instead of this. We'll see.

[smu160]

Closing in favor of #113