On Statistical Inference for High-Dimensional Binary Time Series