RoVi-Aug: Robot and Viewpoint Augmentation for Cross-Embodiment Robot Learning