A manifesto on explainability for artificial intelligence in medicine