ELIGN: Expectation Alignment as a Multi-agent Intrinsic Reward - View it on GitHub