Putting People in Their Place: Affordance-Aware Human Insertion into Scenes