Reinforcement Learning for Continuous Stochastic Control Problems