refactor(llm): return feedbacks as json for better processing and formatting

[fabienzucchet]

Closes issue #22

[github-actions]

LOGAF Level 2 - /home/runner/work/code-review-gpt/code-review-gpt/src/review/prompt/constructPrompt.ts

The appendToLastPrompt function could be improved by using string concatenation instead of creating a new array each time. This would improve performance and readability.

---

LOGAF Level 3 - /home/runner/work/code-review-gpt/code-review-gpt/src/review/llm/feedbackProcessor.ts

1. The collectAndLogFeedback function is not handling the error properly. It's better to let the error propagate up to the caller instead of throwing it again. This way, the caller can decide how to handle the error.

2. The pickBestFeedbacks function could be improved by using a sorting function instead of a priority queue. This would simplify the code and potentially improve performance.

---

LOGAF Level 1 - /home/runner/work/code-review-gpt/code-review-gpt/src/review/llm/AIModel.ts

The apiKey is exposed in the IAIModel interface. This is a security risk. Consider using environment variables to store sensitive information. For example, you can use process.env.API_KEY instead of directly including the key in your code.

---

📝🚀🔒

---

Powered by Code Review GPT

[mattzcarey]

So good