Reinforcement learning and control as probabilistic inference: Tutorial and review