OpenAI Custom http headers

What

This pull request adds the ability to specify custom HTTP headers when integrating with OpenAI via proxy APIs such as Helicone. It also refactors OpenAi's performRequest() by encapsulating the logic for handling the session data task in its own class, URLSessionDataTaskManager. This should simplify the OpenAI class and make the code easier to manage. This pull request simplifies JSONRequest and MultipartFormDataRequest by abstracting out common code to remove repetition that previously existed in these structs. Finally, it simplifies the protocol URLRequestBuildable, moving configuration-driven parameter setting to the constructors of these struct instances, rather than in the call to that protocol's build() function.

Why

The motivation for this change was to allow users to integrate with OpenAI via the Helicone proxy. To do so, they not only need to point the OpenAI client at the Helicone host, but they also have to set an additional HTTP header on the requests. This pull request makes that possible.

Affected Areas

The affected areas of the library include:

The OpenAI class within the OpenAI.swift file.
The client configuration within the constructor of the OpenAI class.
The construction of requests in the following methods: completions(), completionsStream(), images(), embeddings(), chats(), chatsStream(), edits(), model(), models(), moderations(), audioTranscriptions(), and audioTranslations().