Fix empty metadata keyword entries

[floze-the-genius]

Summary

• Drop blank metadata keyword entries produced by trailing or repeated commas.
• Preserve trimming and normal comma-separated keyword behavior.
• Add a focused regression test for the blank-segment case.

Fixes #137.

Tests

• RED: python -m pytest tests/test_generic/test_extractors.py::TestMetadataExtractor::test_extract_keywords_ignores_empty_entries -q returned the old empty entries.
• GREEN: extractor tests — 38 passed.
• GREEN: full suite — 477 passed, 1 skipped, 19 deselected.
• GREEN: ruff check src/ tests/ and ruff format --check src/ tests/.
• GREEN: python -m build.
• GREEN on Python 3.9: extractor tests — 38 passed.

[vedaant00]

Verified locally: the filter is exactly right and the test fails on the old code, so it is a real guard. 38 extractor tests pass, ruff clean. Good to merge, thanks for the clean fix.