Apply one Bellman optimality backup by maximizing expected immediate reward plus discounted next-state value.
