Reward Sharing Based on Past Events in Multiagent Reinforcement Learning Using an External Evaluator