[feature] enable user to parse-in its own OpenAI API key e.g. to choose gpt-4-turbo

[vmeylan]

LLM backend only uses gpt-3.5-turbo-16k without leaving the possibility of gpt-4-turbo which can have 32k or 128k token windows, possibly extending the product capacity.
Would require safely sending the user openai API key to the backend, and the backend accepting such key