Allow multi-deployment configuration for Azure OpenAI

[csotiriou]

Relevant to #237, this will allow overriding the following properties per model in the Azure OpenAI configuration

• endpoint
• resource-name
• deployment-name

Example configuration that is now possible:

quarkus.langchain4j.azure-openai.ad-token=token
quarkus.langchain4j.azure-openai.api-version=2023-07-01-preview
quarkus.langchain4j.azure-openai.chat-model.enabled=true
quarkus.langchain4j.azure-openai.deployment-name=my-deployment-name
quarkus.langchain4j.azure-openai.embedding-model.enabled=true
quarkus.langchain4j.azure-openai.resource-name=name
quarkus.langchain4j.azure-openai.embedding-model.deployment-name=another-deployment-name

In the above example, all models in the parent configuration will have the my-deployment-name, but the Embedding model will have the another-deployment-name specifically as a name.

[geoand]

Very nice!

My only ask is to not use lambdas as we generally try to avoid them in runtime code

[csotiriou]

@geoand I removed the lamdas in favor of Suppliers I hope that's ok (I saw that there were other cases with Suppliers as well. I also made the PR ready for review, as I have tested it and it seems to be working.

I suppose the reason for the change is performance?

[geoand]

Yeah, at some point in the past we saw there was a tiny impact related to lambdas so we decided to try and limit their use in our code

[geoand]

Thanks!

Now all we need is to have the commits squashed 😀

[csotiriou]

Ah, yes. done.

[geoand]

🎉

[geoand]

I force pushed an update that fixes the formatter complaints