Edge Intelligence Optimization for Large Language Model Inference with Batching and Quantization