Use ASCII case folding for HTML element and attribute names

[ac1982]

HTMLTranslator currently Unicode-lowercases element and attribute names. Given <x-Ä id="upper"></x-Ä><x-ä id="lower"></x-ä>, selecting x-Ä incorrectly returns lower. [DATA-Ä] likewise selects a data-ä attribute instead of data-Ä.

Use the existing ascii_lower helper for HTML name normalization, consistent with HTML/CSS ASCII case-matching rules. ASCII portions still fold (X-Ä matches x-Ä), while non-ASCII characters retain their identity. XML/XHTML and the attribute-value extension point are unchanged.

Two regressions fail against master and pass with the fix; they execute the generated XPath against actual lxml-parsed HTML and cover ordinary, mixed-ASCII-case, and wildcard-namespace selectors. This is unrelated to #50 (Python 2 byte-string input), #189/#191 (tokenization), and the recent attribute-value i flag support.

Validation on Python 3.14.4 / Linux arm64:

• Full tests including documentation examples: 29 passed; XPath module 100% coverage, total 99%.
• Ruff lint/format, mypy (5 source files), pylint, complete pre-commit hooks, Sphinx build with -W, sdist build / twine check, and git diff --check passed.
• Other Python versions, PyPy, macOS and Windows were not run locally.

AI assistance: OpenAI Codex investigated, implemented, and ran these checks.