Sampling-Based Optimization for Multi-Agent Model Predictive Control