[Parity #478] Complete learned-model-state diagnostics for every canonical estimator family

Parent: #478

Performance must not outrun semantic evidence. Current benchmark infrastructure has strong state diagnostics for only a subset of rows (notably PCA/KMeans and selected later additions); score parity alone is insufficient for optimizer work.

Add estimator-state comparisons for every canonical family: linear coefficients/intercepts and objective/convergence state; SVM weights/support vectors/dual coefficients/intercepts; tree node structure/thresholds/features/leaves; forest per-tree/bootstrap/RNG structure; GaussianNB priors/means/variances; KMeans centers/inertia/iterations; PCA means/components/singular values; Ridge/Lasso/LinearRegression coefficients/intercepts/objective state; KernelRidge dual coefficients/kernel parameters.

Acceptance: all 19 canonical rows report state_coverage=covered or an explicit semantically justified exception; state deltas are frozen in disparity history; performance classification is blocked when required state evidence regresses beyond contract.

Start by reading parent issue #478 and locating the benchmark infrastructure and canonical-row definitions mentioned in this issue. Trace the existing PCA/KMeans state diagnostics, then determine how each listed estimator family records state coverage and deltas; done means all 19 rows are covered or have justified exceptions, with disparity history and regression blocking updated.