Scoring with Large Language Models: A Study on Measuring Empathy of Responses in Dialogues