More Than the Sum of Its Parts: A Role for the Hippocampus in Configural Reinforcement Learning