Distributionally Robust Multi-Agent Reinforcement Learning for Dynamic Chute Mapping