Tuning Communication Latency for Distributed Model Predictive Control