OpenAI v1 Compatibility in Python API

With the update to v1 OpenAI's API changed significantly, while backwards compatibility was straightforward to preserve on the server the python API is lagging.

The main difference in the pre and pos-v1 APIs is that OpenAI no longer returns dictionary-like objects from and instead fields are accessed directly.

My intention is two preserve the current api of the `Llama` class and not introduce any breaking changes however it would be nice to support the new API in such a way that OpenAI applications can swap out and use a local llama.cpp model.