Learning Locally Interacting Discrete Dynamical Systems: Towards Data-Efficient and Scalable Prediction