Document HuggingFace Integration

[GregoryComer]

📚 The doc issue

With the recent work on HuggingFace integration for ExecuTorch, we want to highlight this as part of the main ET docs so that users know to use HF for LLMs and how to integrate end to end from C++, Java, or Obj-C/Swift, ideally with minimal code.

Within the new doc flow, I'm tentatively thinking we create a top-level HF page under Usage/, maybe "Using Models from HuggingFace"? This can cover all of the steps needed to export the model and run on device. Ideally we can provide provide inline example snippets and an end-to-end example for a common transformer model.

We can also add a callout and link in Getting Started with ExecuTorch under the Preparing the Model section and in Model Export and Lowering. There is a Working with LLMs section, but I'd like to overhaul this section completely, so let's maybe hold off on that area for now.

On the runtime side, I'd like to make sure there is a clear path for runtime integration on Android and iOS, in particular, using Java and Obj-C/Swift. Ideally we define the "core" LLM API surface and components, document them, and provide API parity across all language bindings. Tracking in #8767.

Suggest a potential alternative/fix

No response

cc @mergennachin @byjlw @cccclai @helunwencser @jackzhxng