refactor: openai-clone providers

[0xMochan]

• I have looked for existing issues (including closed) about this

Feature Request

• Split up openai into multiple modules
• Refactor openai to make reuse much simpler and easier for alternative clients

Motivation

• Some providers such as openai have grown fairly complex and should be split up into multiple files.
• Other providers which are based on openai have varying amount of code duplication which leads to bug duplication with replicating fixes and improvements, esp. to openai.

Note

• Perhaps openai can have provider traits?
• Reusing Message is usually the easiest amongst the openai compatible
• Reusing CompletionRequest is the best but unreasonable for those who have excess data
  • Perhaps, #[serde(flatten)] can be used for superset?
• The base completion implementation needs to refer to some standard parsing structure for atleast the message handling
  • Helper method in openai module?
• Consider multiple PRs: 1 for splitting up openai, another for each provider!