fix(etag): reject codepoints above 0xFF in ETag.parse

[koteshyelamati]

What

Adds the missing upper-bound check to ETag.parse's character-set guard so codepoints above 0xFF are rejected, matching RFC 7232 §2.3 (etagc = %x21 / %x23-7E / obs-text, obs-text = %x80-FF) and the inline comment that already documents this boundary. Without the bound, any Python str codepoint at or above 0x80 was accepted, including codepoints above 0xFF (e.g. €) that fall outside obs-text and are not valid entity-tag characters. The accepted value then round-trips through ETag.__str__, which request_conditions._format_etags uses when building If-Match / If-None-Match headers, so a non-conformant tag could flow onto the wire instead of being caught at parse time.

Also adds test_parse_above_obs_text_raises, pinning rejection of a codepoint above 0xFF alongside the existing control-char and space cases in tests/http/test_etag.py.

Fixes #19

Checklist

• uv run pytest -q passes
• uv run mypy --strict passes
• uv run ruff check and uv run ruff format --check pass
• New/changed public symbols have Google-style docstrings and appear in __all__
• New .py files carry the two-line MIT licence header
• Commit messages use the feat: / fix: / chore: / docs: / ci: prefixes

Notes:

• This is a one-line logic fix plus a one-line test addition; no new public symbols or new files, so the docstring/licence-header items don't apply.
• I edited via GitHub's web UI and didn't have a local checkout to run pytest/mypy/ruff myself, so I've left those unchecked for CI to confirm. The fix matches the exact change suggested in the issue.
• The second commit uses a test: prefix, which isn't in the list above (feat/fix/chore/docs/ci). Happy to squash into a single fix: commit if you'd prefer.