fix bf16 fneg crash on AArch64 (Apple M4)

[rickwoi]

fix #133.

the bitwise fallback for -(::BFloat16) (using xor) is optimized to fneg by the compiler.

this causes an LLVM "Cannot select" error on AArch64 (Apple M4) which lacks the (fneg bf16) rule.

changing the fallback to BFloat16(0) - x forces fsub instead, which avoids the crash.

also added regression tests, confirmed to pass on CI (local Termux env can't run suite).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.01%. Comparing base (bd1ff36) to head (5472b8f).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #114   +/-   ##
=======================================
  Coverage   97.01%   97.01%           
=======================================
  Files           3        3           
  Lines         201      201           
=======================================
  Hits          195      195           
  Misses          6        6           

☔ 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.

[rickwoi]

sorry, for this #113