CatBoost-like Numeric Estimators

MPSBoost provides CatBoost-like numeric estimator entries:

• CatBoostRegressor
• CatBoostClassifier

The current focus is an ordered-boosting-style sklearn-compatible public API and numeric feature workflow.

Supported capabilities:

• ordered boosting permutation semantics
• numeric feature training
• categorical feature parameter compatibility entries
• shared native tree engine
• classifier / regressor API
• sample weight
• regularization controls

Limitations:

• Categorical model persistence remains future model-format work.
• MPSBoost does not call CatBoost as a hidden training engine.