Fairness Meets Privacy: Integrating Differential Privacy and Demographic Parity in Multi-class Classification