Module step 

Available on crate feature alloc only.

Single boosting step: owns one tree + drift detector + optional alternate.

BoostingStep is a thin wrapper around TreeSlot that adds SGBT variant logic. The three variants from Gunasekara et al. (2024) are:

• Standard (train_count = 1): each sample trains the tree exactly once.
• Skip (train_count = 0): the sample is skipped (only prediction returned).
• Multiple Iterations (train_count > 1): the sample trains the tree multiple times, weighted by the hessian.

The variant logic is computed externally (by SGBTVariant::train_count() in the ensemble orchestrator) and passed in as train_count. This keeps BoostingStep focused on execution rather than policy.

Structs

BoostingStep

A single step in the SGBT boosting sequence.