Stop fit() mutating constructor parameters in PCA / PLS / OPLS (#505)

[kgdunn]

Summary

• Fixes PCA / PLS / OPLS mutate constructor parameters inside fit(), breaking the sklearn clone contract that cross_validate depends on #505: fit() no longer writes the resolved component count (or, in PLS, the resolved missing-data settings) back onto constructor parameters. The resolved value is the new fitted attribute n_components_; n_components stays exactly what the user passed, including None, so get_params/clone/PLS.cross_validate resamples fit the requested configuration and refitting one instance on differently shaped data re-derives the clamp each time. Clamping logic and its SpecificationWarning are unchanged, per the issue's non-goals.
• Internal post-fit readers migrated to n_components_: the shared Hotelling's T2 limit mixin and ellipse_coordinates in _base.py, spe_limit in _limits.py, the plot pre-checks in plots.py (new _fitted_n_components helper that also handles the TPLS _parent wiring), PCA's hotellings_t2, and PLS prediction_interval. MBPCA/MBPLS now also set n_components_ in fit() so those shared helpers read one attribute across all four estimators.
• OPLS (which has no n_components constructor parameter) exposes the fitted count only as n_components_ and maps the old attribute name to a helpful rename message via its existing _ATTRIBUTE_RENAMES machinery.

All four acceptance criteria from the issue are covered by new tests in tests/test_sklearn_compat.py: get_params() round-trip after fit (including None), clone() reproduces the request, refit on different shapes resolves independently, and a spy on PLS.fit pins that every cross_validate sub-model is constructed with the user's requested n_components.

Version 1.73.0 (MINOR: post-fit meaning of n_components changes; assumes queued PR #524 merges first, trivial rebase otherwise).

Test plan

• New PCA / PLS / OPLS mutate constructor parameters inside fit(), breaking the sklearn clone contract that cross_validate depends on #505 acceptance tests in tests/test_sklearn_compat.py (20 passed)
• uv run pytest tests/test_multivariate.py tests/test_multivariate_opls.py tests/test_sklearn_compat.py tests/test_multivariate_contribution_plots.py tests/test_audit_regressions_multivariate.py --no-cov: 269 passed, 2 pre-existing skips
• uv run pytest tests/test_multiblock_reference.py tests/test_multiblock_oracles.py tests/test_multivariate_tpls_display.py tests/test_multivariate_adaptive.py tests/test_multivariate_tpls_feature_importance.py --no-cov: 154 passed
• ruff check and ruff format --check clean; mypy src/process_improve shows only the pre-existing mcp_server.py FastMCP error unrelated to this diff

Checklist

• Version bumped in pyproject.toml (PATCH for fixes/docs/config, MINOR for new features)
• Tests added or updated where relevant
• ruff check . passes
• CHANGELOG.md updated

---

Generated by Claude Code

[kgdunn]

The typecheck failure here is not this PR's: it is red on main's own latest runs. The unpinned mcp extra now resolves to mcp 2.x, where FastMCP was renamed MCPServer, so mypy fails on src/process_improve/mcp_server.py everywhere. The fix is PR #528; its migrated mcp_server.py and mcp>=2.0 floor are ported verbatim into this branch so CI can go green in any merge order. The ported files disappear from this diff once #528 merges.

---

Generated by Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 93.33333% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/process_improve/multivariate/plots.py	87.50%	1 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!