Incentive-Compatible Federated Learning with Stackelberg Game Modeling