Human Gaze Boosts Object-Centered Representation Learning