Detecting LLM Fact-conflicting Hallucinations Enhanced by Temporal-logic-based Reasoning