Respect custom separators for C/POSIX locale aliases

[PDGGK]

Summary

Respect sep when normalizing the C/POSIX locale aliases.

Locale.parse("C", sep="-") and Locale.parse("POSIX", sep="-") currently fail because alias normalization builds en_US_POSIX, which is then parsed using -. The explicit canonical input Locale.parse("en-US-POSIX", sep="-") succeeds.

Build the alias with sep.join(("en", "US", "POSIX")) instead. This preserves the default underscore behavior and allows the supported hyphen separator. Public Locale.negotiate(["C"], ["C"], sep="-") follows the repaired path too.

This is a small follow-up to #1340. It does not change encoding/modifier handling or make C/POSIX and their canonical spellings interchangeable availability entries during negotiation.

Validation

Linux, CPython 3.12.14:

• 16 focused cases pass; original source and production-only revert reproduce 9 failures with 7 passing controls.
• Full suite with normal doctest collection: 7,846 passed, 7 skipped, 2 expected failures.
• Pinned Ruff and all pre-commit hooks pass.
• Source distribution/wheel builds and clean-wheel parsing, negotiation, and formatting smoke checks pass.

The skips and expected failures are pre-existing. Windows, macOS, other supported Python versions, and hosted CI have not been run for this patch.

AI assistance: implementation, tests, and this description were prepared with OpenAI assistance.