Add support for GPT-4 Vision

[ronaldmannak]

This PR adds support for GPT-4 Vision. Images can be added to a message as a URL or as Data.

The token counter for small images adds the correct amount of tokens. However, large images have a complex way of counting tokens and require the library to download the image. To keep things simple, this PR just adds the maximum number of tokens possible. Maybe a feature could be added in the future that optionally downloads the images and counts the tokens the right way.

let media = [
  MessageContent.text("Count the number of apples in this image"),   
  MessageContent.imageUrl("https://apples.com/apples.jpg"),
 ]
thread.addUserMessage(_ media: media) 

or

let imageData = UIImage(named: "apples.jpg").jpegData(compressionQuality: 1.0) 
let media = [
  MessageContent.text("Count the number of apples in this image"), 
  MessageContent.imageUrl(URLDetails(imageData: imageData))]
thread.addUserMessage(_ media: media) 

There is one breaking change. The content property in ChatMessage is now an enum and invoking description might be needed to fetch the string, as was the case in one of the unit tests:

  XCTAssertEqual(userMessageContent, chatThread.getNonSystemMessages().first?.content) // old
  XCTAssertEqual(userMessageContent, chatThread.getNonSystemMessages().first?.content?.description) // new