CV Predict Classifier

Classifies feature rows with a PREVIOUSLY TRAINED cv2.ml model: pick a file from ComfyUI/models/cvml - the gzip '.yml.gz' files 'CV Save Classifier' writes, or a plain '.yml'/'.yaml' dropped in by hand - or connect a model_yaml STRING to skip the file entirely (it takes precedence when connected). Loading and predicting happen inside one call - the model never crosses a graph edge, so caching stays correct (the same pure-function argument as 'CV Train Classifier'). The features must have the SAME length D the model was trained on (e.g. the same 'CV HOG Features' settings). Predictions are the label values the model was trained on - 'CV Train Classifier' trains on class INDICES 0..K-1 in 'CV Stack Feature Classes' input order (class_a = 0, ...). Failure-tolerant: a missing/corrupt file, a non-model YAML, a feature-length mismatch or empty features yield success=false and empty predictions, never an error - gate downstream consumers with if/else.