Adaptive Client Sampling in Federated Learning via Online Learning with Bandit Feedback