Large-Scale Mixed-Traffic and Intersection Control using Multi-agent Reinforcement Learning