fix(quantization): use raw docstrings for '\*' escapes

[alphasafal]

The OpQuantizerConfig and ModuleQuantizerConfig docstrings escape * as \* so Sphinx doesn't treat it as emphasis. In a normal string literal that's an invalid escape sequence, so on Python 3.12+ compiling quantization_config.py prints

SyntaxWarning: invalid escape sequence '\*'

and with python -W error (or a pytest config that turns warnings into errors) the module doesn't compile at all. I ran into it right after installing on 3.13.

Changes

• Make both docstrings raw. The string values are identical to before (the backslash is kept either way), so the rendered docs don't change. I compared the AST before and after to be sure.
• Same r prefix on four regex literals in test_annotation_pattern_registry.py that had \. / \*.
• Enable ruff W605 so make check catches new ones. On main it flags exactly these 20 spots.
• Add tests/test_source_compiles.py, which compiles every module under coreai_opt with warnings as errors. It fails on main and passes with this change.
• Changelog fragment under Fixed.

Testing

• pre-commit run --all-files passes.
• pytest -m "not slow" on Python 3.13 / Linux: everything passes apart from export and nox tests that need coremltools, pyreporoot and model downloads I couldn't install in my environment. None of those touch the files changed here.