Improve azure openai support

[BramVanroy]

• Adds instructions how to use OpenAI services through Azure in README
• Adds a minor typing improvement for an Azure client
• Skip model checking for Azure models because the Azure client is a weird thing: instead of a model like "gpt4-0123" it requires the deployment name. So self.model can be any arbitrary string, so verifying it against supported models won't work.

[alvarobartt]

Hi @BramVanroy thanks for this! We're about to close the refactor in the core-refactor branch, everything's new and refactored there, so we can either merge yours in main and adapt that ourselves to the refactored version, or you can do that yourself, whatever you feel is better 👍🏻 It should be straight forward to extend the OpenAILLM class to create a subclass for AzureOpenAILLM, but ping us if you have any doubt and willing to work on top of core-refactor, otherwise, we'll be happy to review this one! Thanks again

[BramVanroy]

Hi @alvarobartt I do not have the time to work on this so preferably the path that takes the least effort from my side. :)

If you plan to write a subclass for OpenAI specifically for Azure, that could be even better because then users can pass in their api_version, endpoint, deployment name, etc. rather than have to pre-define the azure client.