Deployable Reinforcement Learning with Variable Control Rate