Exchanging Lessons Between Algorithmic Fairness and Domain Generalization