v0.1.2

What's Changed

• fix: 3 data-leakage/correctness bugs + FDR-correct Boruta [skip-doc-check] by @SeanPeppers in #86
• feat: collinearity gates the Drop decision [skip-doc-check] by @SeanPeppers in #87
• docs: stop calling confidence scores 'calibrated' by @SeanPeppers in #88
• chore: pipeline code-quality cleanup (dead code, duplication, tests) [skip-doc-check] by @SeanPeppers in #89
• fix: collinearity gating's reason text overclaimed 'Caution' as final [skip-doc-check] by @SeanPeppers in #90
• chore: benchmark Keep/Drop recommendations against ground-truth datasets [skip-doc-check] by @SeanPeppers in #91
• feat: shadow probes gate the Drop decision [skip-doc-check] by @SeanPeppers in #92
• docs: honest rationale for the unvalidated magic-number thresholds [skip-doc-check] by @SeanPeppers in #94
• [skip-doc-check] Build/test/review Rust LSH extension by @SeanPeppers in #93
• fix: consolidate synthesize_features auto-fix into Config [skip-doc-check] by @SeanPeppers in #95
• test: real MLflow tracking store test + verify/fix Sphinx docs build [skip-doc-check] by @SeanPeppers in #96
• [skip-doc-check] fix: auto-enable shadow-probe noise floor gating under severe class imbalance by @SeanPeppers in #97
• feat: collapse feature groups into aggregate Keep/Caution/Drop rows by @SeanPeppers in #98
• feat: token-level SHAP attribution for text/transformer models by @SeanPeppers in #99
• ci: add CodeQL and dependency review, use private vulnerability reporting by @SeanPeppers in #124
• fix(deps): upgrade locked dependencies to clear known advisories by @SeanPeppers in #125
• ci: install from uv.lock everywhere and add a hypothesis fuzz test by @SeanPeppers in #126
• ci: sign release artifacts with Sigstore, bump to 0.1.2 by @SeanPeppers in #127

Full Changelog: v0.1.1...v0.1.2