Logic-of-Thought: Injecting Logic into Contexts for Full Reasoning in Large Language Models