Interpreting Robustness Proofs of Deep Neural Networks