Goal Recognition as Reinforcement Learning