Reinforcement Learning with Lookahead Information