Assessing Large Language Models for Automated Feedback Generation in Learning Programming Problem Solving