refactor(tree,ensemble): address review follow-ups from #467

[Mec-iS]

Follow-up to #467. Applies the three inline notes from the review.

Checklist

• My branch is up-to-date with main branch.
• Everything works and tested on latest stable Rust.
• Coverage and Linting have been applied

Current behaviour

Three review notes on #467 stayed open: find_best_cutoff recomputes the node mass on every entry (one O(n) pass per node); a test comment says "No bootstrapping" where bootstrap: true is set; validate_sample_weights is pub(crate) in base_tree_regressor, so an internal helper leaks into three modules.

New expected behaviour

• mass is passed down like the other split statistics: fit_weak_learner gives the root mass to find_best_cutoff; split accumulates the child masses in the existing partition loop and hands them to the recursive find_best_cutoff calls. The re-computation block is removed.
• The test comment now reads "Use bootstrapping".
• validate_sample_weights is private to each entry-point module (decision_tree_regressor, random_forest_regressor, extra_trees_regressor). No shared helper crosses module bounds.

Numerical behaviour is unchanged: same values and summation order on every path. cargo test --all-features passes (554 unit tests + doctests), cargo clippy --all-features -- -Drust-2018-idioms -Drust-2024-compatibility -Dwarnings and cargo fmt --all -- --check are clean.

Change logs

Changed

• tree: BaseTreeRegressor::find_best_cutoff takes the node mass as a parameter. The root mass comes from fit_weak_learner; child masses are accumulated in split during the sample partition. Internal API only.
• tree, ensemble: validate_sample_weights is private to each of the three entry-point modules instead of pub(crate) in base_tree_regressor. Internal API only.