From Robustness to Explainability and Back Again