The Optimal Reward Operator in Dynamic Programming