Evaluating individualized treatment effect predictions: a new perspective on discrimination and calibration assessment