Reinforcement Learning for Control: Performance, Stability, and Deep Approximators