FedCal: Achieving Local and Global Calibration in Federated Learning via Aggregated Parameterized Scaler