Articulating user preferences in many-objective problems by sampling the weighted hypervolume