fix crash on aarch64 linux wheels

[davidhewitt]

Investigating #174

[codspeed-hq]

CodSpeed Performance Report

Merging #175 will not alter performance

_{Comparing dh/aarch64-ci-test (1cbf565) with main (4397293)}

Summary

✅ 73 untouched benchmarks

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.75%. Comparing base (4397293) to head (1cbf565).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #175      +/-   ##
==========================================
+ Coverage   88.74%   88.75%   +0.01%     
==========================================
  Files          13       13              
  Lines        2203     2206       +3     
  Branches     2203     2206       +3     
==========================================
+ Hits         1955     1958       +3     
  Misses        153      153              
  Partials       95       95              

Files with missing lines	Coverage Δ
crates/jiter/src/py_string_cache.rs	93.16% <100.00%> (+0.17%)	⬆️

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4397293...1cbf565. Read the comment docs.

[davidhewitt]

@samuelcolvin looks like pystring_ascii_new is broken on aarch64 linux. That's... not a huge surprise, the underlying FFI there is extremely suspect, it's the worst bit of pyo3-ffi by far. I can't see anything obviously wrong but the bitflags stuff inside PyUnicode_DATA has always reeked of possible unsoundness. I'll follow up on PyO3 but for now disabling here.

I've disabled this on everything except x86_64 linux and arm macOS as the most well-tested platforms, to be safe.