Imitation and Reinforcement Learning with Heterogeneous Actions