Sanitize Content-Disposition filename to a safe basename

[Flerpharos]

Fixes #326

src/labapi/entry/entries/attachment.py used the server-supplied Content-Disposition filename (msg.get_filename()) verbatim, while the S3-fallback path (_s3_filename_from_url) already reduced its filename to a sanitized basename. A download response with a Content-Disposition filename such as ../../x could therefore leave Attachment.filename carrying path-traversal sequences.

Fix

• Factored the sanitization logic out of _s3_filename_from_url into a shared _safe_basename() helper: reduces to PurePosixPath(name).name, rejects None/empty/./.., and normalizes backslashes to forward slashes first so Windows-style traversal sequences (..\..\x) are also handled.
• _ensure_attachment now applies _safe_basename() to msg.get_filename() before it becomes Attachment.filename.
• The page-listing filename (self._filename, supplied at upload time) is left as-is — unrelated to this server-response parsing path, and already covered by an existing test asserting it takes precedence unchanged.

Tests

Added test_attachment_entry_sanitizes_content_disposition_filename (parametrized) in tests/entry/entries/test_attachment.py covering:

• ../../x → x
• ..\..\x (backslash variant) → x
• a normal filename (report.pdf) is unchanged

Verification

• uv run pytest --ignore=tests/test_integration.py -q — 433 passed, 1 skipped
• uv run ruff check . — all checks passed
• uv run ruff format --check . — all files formatted