Understanding the Uncertainty of LLM Explanations: A Perspective Based on Reasoning Topology