Input_training(cross_validation = TRUE) with
current xgboost releases by replacing the
caret::train(method = "xgbTree") tuning path with native
xgboost::xgb.cv() while preserving the
MethScopeModel return structure used by
PredictCellType().