feat: add bivariate Shape²CART branching

[upadhyan]

Summary

• add opt-in Shape²CART bivariate routing to classifiers, regressors, and random SG forests while preserving the univariate default
• support binary and multiway pair routers across continuous, categorical, mixed, and per-feature missing-value cases
• make TAO pair-aware, export/replay pair routers, and reject impurity feature importances after TAO refinement
• render exact pair-routing heatmaps with marginal histograms, branch-changing threshold labels, categorical labels, and missing-value margins
• add the executed S²GT tutorial and update the related quickstart, API, roadmap, and cross-document guidance

Validation

• 1214 passed, 97 skipped
• strict Sphinx build: python -m sphinx -W --keep-going -b html docs docs/_build/html
• Ruff passes for the changed Python and test files
• git diff --check

Refs #27
Closes #28
Closes #42
Closes #43
Closes #44
Closes #45
Closes #46
Closes #47
Closes #48
Closes #49