Multi-Objective Optimization for Privacy-Utility Balance in Differentially Private Federated Learning