Stale PR data issue

[shemminger]

If I run codedog python script once to review a github PR with openai then use it again to review a different PR it produces the old report on the second run. Some part of the data is being cached locally or by OpenAI and therefore getting same report twice.

Using gpt-4.1-mini as the model since full gpt exceeds usage limits.