Adding support for GPT-4o and GPT-4-turbo.

[wz-ml]

Changes:

• Added support for GPT-4-turbo (🚀) and GPT-4o (🅾️).
• Removed support for instruct models and added back streaming.

[giosilvi]

Starting the review, so far it looks good.

[giosilvi]

This needs to be export const CHAT_API_MODELS to work correctly. And the imports should become:
import { CHAT_API_MODELS } from.
All the import were not actually working, so all the if model in CHAT_API_MODELS conditions (apart from gtp3.js) were false by default.
Note also, that now all prompt should have the prompt formatted as:
{ "role": "user", "content": text }
including the custom prompts in background.js.

Given that now all models are chat-type we may add a switch to chose for the fast prompt or chat-gpt type of prompt

[giosilvi]

I pushed a fix for the above. This should be enough unless I missed something else.