CLAIR-A: Leveraging Large Language Models to Judge Audio Captions