MAMPS: Safe Multi-Agent Reinforcement Learning via Model Predictive Shielding | Read Paper on Bytez