Introduce Vision API

[belaszalontai]

Based on this doc https://platform.openai.com/docs/guides/vision without proper API reference

[belaszalontai]

@kayhantolga PLease review. Unfortunately so far there is no clear API reference for Vision feature at open AI's docs. Only a general guide exists.

Here are some confusing point:

• the content of a Message in messages is different for ChatCompletion with normal models and chat completion with vision model. For normal models it is a string type, for vision model it is a json array type with more content. Maybe it could be used mixed but I decided in this implementation to implement only the json array for vision model.
• FromAssistant (Assistant role) is available but for what?
• should I implement some helper for base64 encoding? If yes where should I place it in the code?

[belaszalontai]

meanwhile I have realized that there is an API description for chat completion which defines the type of content for Vision API here: https://platform.openai.com/docs/api-reference/chat/create

Acoording to this reference, the

• System, Assistant message can only be string
• User message can be string or array

So I am refactoring this PR..

[hucongquan]

Hello， 你发的邮件我已经收到啦，Thx！

[belaszalontai]

New PR will be placed soon