v1.1.0: Scale fuzzy matching to 50K rows, version bump

[MsShawnP]

Summary

• Version bumped to 1.1.0 — aligns with PyPI release tag and CHANGELOG
• Fuzzy matching scales to 50,000 rows — n-gram blocking replaces the old 500-row brute-force cap. Character trigram inverted index generates candidate pairs, then Levenshtein only runs on plausible matches
• Small datasets (≤500 rows) still use brute-force (faster for them)

How n-gram blocking works

1. Generate character 3-grams for each normalized record
2. Build inverted index: trigram → list of row indices
3. Count shared trigrams per pair, skip very common trigrams (noise)
4. Only run Levenshtein on pairs sharing enough trigrams
5. Cap candidates per record (50) to bound worst-case

Test plan

• 220 tests pass (8 new)
• ruff check . — clean
• mypy data_hygiene_auditor/ — 0 errors
• N-gram blocking unit tests (similar/dissimilar/empty/short/1000-row scale)
• Large dataset test: 600 rows with injected typo pair found without skip warning
• 501-row test confirms old limit no longer triggers skip

🤖 Generated with Claude Code