API: preprocessing.feature_selection

skyulf.preprocessing.feature_selection

Feature-selection nodes (Variance / Correlation / Univariate / Model-based).

Each Applier dispatches on engine via :func:apply_dual_engine (see dispatcher.py); per-engine logic lives in small _apply_polars / _apply_pandas static helpers. Calculator fit paths are sklearn-bound, so they use :func:to_pandas once at the top instead of going through fit_dual_engine.