Distributed Online Learning for Joint Regret with Communication Constraints