Addressing Client Drift in Federated Continual Learning with Adaptive Optimization