Batch phase 3: sklearn transformer facades and f_rupture

[kgdunn]

Summary

• Adds sklearn BaseEstimator / TransformerMixin facades over the existing batch free functions, so a batch workflow can be expressed as a pipeline and the learned state is carried on the fitted estimator: BatchScaler (range scaling with inverse_transform), ResampleAligner (linear resampling to a common length), DTWAligner (iterative weighted DTW that learns a reference and weights and aligns new batches to them), and BatchFeatureExtractor (a batch dict to a batch-by-feature matrix suitable as the X block of a PLS-to-quality model). The free functions are unchanged and remain the implementation layer.
• Implements the previously-stubbed f_rupture via the optional ruptures library (PELT changepoint detection), and re-points the batch extra to ruptures - it previously declared openpyxl and scikit-image, which were imported nowhere.

This is Phase 3 of the batch modernization plan (pipeline ergonomics + feature completion).

Note on base branch: stacked on claude/batch-phase2-online-monitoring (#460), which is stacked on Phase 1 (#459). GitHub will retarget the base automatically as each parent merges. Review #459, then #460, then this.

Test plan

• tests/batch/test_transformers.py: scaler round-trip; resampler equal-length + auto-reference; DTW aligns the training set and a held-out batch to the reference length; feature matrix shape/index and feeding a PLS model; unknown-feature guard; f_rupture detects an injected step (guarded by importorskip("ruptures")) and rejects multi-column input
• Full batch suite: 121 passed; ruff check . and mypy src/process_improve clean

Checklist

• Version bumped in pyproject.toml (MINOR: 1.56.0 -> 1.57.0) and CITATION.cff synced
• Tests added or updated where relevant
• ruff check . passes
• CHANGELOG.md updated

🤖 Generated with Claude Code

https://claude.ai/code/session_01SyEpexCyqHn1rwTMYQSiP6

---

Generated by Claude Code