Optimal Execution with Reinforcement Learning