Sequential Decision Making with Rank Dependent Utility: A Minimax Regret Approach