FORESEE: Multimodal and Multi-view Representation Learning for Robust Prediction of Cancer Survival