Sim-and-Real Reinforcement Learning for Manipulation: A Consensus-based Approach