LLM-QBench: A Benchmark Towards the Best Practice for Post-training Quantization of Large Language Models