Include x-coding-assistant=open-interpreter header in litellm calls

[jhrozek]

Describe the changes you have made:

Proxies that inspect traffic between the development environment and an
LLM might be interested in whether it's OI or another tool calling in
order to be able to inspect and/or modify the payload.

The most common way of solving this would be to add a user-agent header.
However, litellm which OI uses calls into OpenAI libraries directly
when making the request and it seems like the only way to set a custom
http_client. This seemed like something that might have unforeseen
consequences (timeouts? retries?). For other LLMs, litellm seems to use
its own httpx wrapper which might presumably be easier to customize, but
I have not tried.

To make things easier, let's just add an OI specific header. I put the
string OI followed by the version there, but the value - and indeed the key

• of the header are not that interesting, what I would like to do is to just
  be able to to tell OI calls.

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

N/A

Pre-Submission Checklist (optional but appreciated):

• I have included relevant documentation updates (stored in /docs)
• 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

[jhrozek]

@KillianLucas is this something that OpenInterpreter would be interested in merging?