Toward Automated Qualitative Analysis: Leveraging Large Language Models for Tutoring Dialogue Evaluation