Marrying Fairness and Explainability in Supervised Learning