Add a "cost" attribute to LLMContent response, fix bug with backup model, fix timeout

[rishsriv]

Features Added

A new "cost" parameter to the response of LLMContent. This allows us to track how much an LLM call costs, and will help us down the line when we start to track all of our LLM calls.

Bugs fixed

• Fixed an issue with the new backup_model parameter to make it work
• Moved timeouts to function level instead of client level calls, as client levels calls weren't working for the OpenAI client