Scalable Coordinated Exploration in Concurrent Reinforcement Learning