On the explainable properties of 1-Lipschitz Neural Networks: An Optimal Transport Perspective