OpenAI: Support audio modality in chat completion

[ThomasVitale]

OpenAI has finally launched support for audio modality in its chat completion API, both as input and as output.

We should start by supporting the input audio modality, in line with the existing APIs in Spring AI for multimodality.
The output audio modality can be supported at a lower level (OpenAiApi), but it will need some design discussions to surface it through the Spring AI abstractions.

More information: https://platform.openai.com/docs/guides/audio

[ThomasVitale]

Here's a PR: #1561