API: preprocessing.resampling

skyulf.preprocessing.resampling

Resampling nodes (Oversampling, Undersampling).

Both Appliers route through :func:apply_dual_engine. Imblearn is purely pandas/numpy-bound, so the Polars path round-trips through pandas (convert in, run sampler, convert back) while keeping all engine handling out of class bodies. Sampler construction is split per-method and per-family so each helper stays at low CCN.