Decode extended filenames for streamed downloads

[saitakarcesme]

Downloads with a Content-Disposition filename* parameter currently fail to infer a filename. Decode the declared charset and percent escapes, prefer a valid extended filename over the ordinary fallback regardless of order, and strip directory components after decoding. Malformed encodings fall back to filename.

Fixes #379. Follows RFC 6266 section 4.3.

Added coverage for UTF-8, ISO-8859-1, language tags, parameter casing, precedence, malformed values, encoded path separators/drive prefixes, and writing the downloaded content to the inferred Unicode filename. Nine regressions failed before implementation.

Validation:

• Python 3.13: 184 passed, 5 skipped.
• Python 3.11 with pyOpenSSL 24.0.0: 184 passed, 5 skipped with deprecation warnings treated as errors. Newer pyOpenSSL triggers existing X509 adapter deprecations under that strict setting.
• Package flake8 and git diff --check pass. The test file retains two pre-existing style findings on unchanged lines.
• Sphinx HTML build succeeds with an existing theme deprecation warning.
• Python 2.7 is not available locally; the existing compatibility style is retained but that interpreter was not tested.

Disclosure: implemented and tested using OpenAI Codex in an authorized automated contribution session.