Cancer drug response prediction with special focus on ML approaches: explainability, compositional generalization, integrating biological knowledge, uncertainty, domain adaptation, few shot learning, quick/fast adaptation, test time adaptation, drug representations that interact with tumor expression profiles