Improved litellm exception handler

[TheFirstRandom]

Describe the changes you have made:

I improved the litellm exception handler in llm.py by adding a 15 sec backoff, if litellm returns litellm.exceptions.RateLimitError. If the rate limit exceeds a second time, interpreter displays a message that you should try again later and check your balance if you use an API. RateLimitError can be caused by depleted tokens.

Reference any relevant issues (e.g. "Fixes #000"):

None

Pre-Submission Checklist (optional but appreciated):

• I have included relevant documentation updates (stored in /docs) -> Think there are no changes to make at all
• I have read docs/CONTRIBUTING.md
• I have read docs/ROADMAP.md

OS Tests (optional but appreciated):

• Tested on Windows
• Tested on MacOS
• Tested on Linux