Exploring the Potential of Anomaly Detection Through Reasoning with Large Language Models