Can ChatGPT rate body-related images like humans? Effects of temperature and few-shot prompting on valence and arousal judgments