use PySys_WriteStderr instead of eprintln

[Person-93]

This is part of the WIP no_std support.

There are two places that were using the eprintln macro, this replaces them with a safe wrapper around PySys_WriteStderr.

This is arguably not a purely internal change. If this PR is merged, those error messages will be redirected (along with all other error output) if the user redirects Python's stderr. Maintainers, please let me know if this needs a news fragment and if it does, if it should be a fix or a change.

[Tpt]

Maintainers, please let me know if this needs a news fragment and if it does, if it should be a fix or a change.

Yes, would be great! A "change" looks like the most appropriate to me but I do not have a strong opinion

[Person-93]

After reviewing Icxolu's comment, I don't think calling python API to write to stderr is valid during GC traversal, so I changed it to use libc::write.

[codspeed-hq]

Merging this PR will not alter performance

✅ 141 untouched benchmarks
⏩ 6 skipped benchmarks¹

---

_{Comparing Person-93:no_std_err_output (df3c3d9) with main (a030369)}

Footnotes

1. 6 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[Person-93]

Is there anything blocking this? Is using libc::write until #6330 merges ok?