Extracting Interpretable Models from Tree Ensembles: Computational and Statistical Perspectives