Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment