fix(download): parse quoted and extended response filenames

[KennyMcSimpson]

The download runner treats a response filename as a regular-expression match, which truncates quoted names such as filename="paper;v2.pdf". It also does not handle filename* correctly, so a server-provided Unicode filename can be lost or replaced by the fallback URL name.

Use the standard email header parser for quoted parameters and RFC 2231/5987 extended filenames. Prefer the extended filename when both forms are present, and leave literal percent sequences in ordinary filename values intact. The result still goes through the existing basename sanitization.

Validation:

• pytest tests/test_download_filenames.py -q: 20 passed, covering quoted punctuation, extended filenames, parameter order, URL fallback, explicit-name precedence, and encoded path separators.
• The CI Ruff command and the new test file pass lint.
• Earlier full-suite run on this exact source tree: 4,798 passed, 5 skipped; the redirect-validation test fails because the container cannot resolve www.iana.org. That test and its fetch implementation are unchanged.
• Pyright and import smoke passed. The external API preflight could not complete without a valid API key.