The Beta Policy for Continuous Control Reinforcement Learning