Upper-contour strategy-proofness in the random assignment problem