Towards User Scheduling for 6G: A Fairness-Oriented Scheduler Using Multi-Agent Reinforcement Learning