Autoformalizing and Simulating Game-Theoretic Scenarios using LLM-augmented Agents