fix: compile UTF-8 parser with unsigned chars

[darvid]

Summary

• build Vectorscan/Hyperscan with unsigned char semantics on non-Windows targets
• add issue Compile fails with "Expression is not valid UTF-8" when using UTF-8 and HS_FLAG_UTF8 #274 regression coverage for HS_FLAG_UTF8 with multibyte UTF-8 literals
• stop skipping UTF-8 validation failures in the unicode test

Verification

• uv run pytest tests/test_hyperscan.py::test_unicode_expressions -vvv
• uv run --with pytest-mock pytest tests/ -vvv
• reproduced published hyperscan==0.9.0 failure in linux/x86_64 Docker
• verified patched source build in linux/x86_64 Docker

Fixes #274