The Sample Complexity Of Simple Reinforcement Learning