LIBRE: Learning interpretable boolean rule ensembles