DM$^2$: Decentralized Multi-Agent Reinforcement Learning for Distribution Matching