Revamp export llm documentation

📚 The doc issue

https://docs.pytorch.org/executorch/1.0/llm/getting-started.html

Currently we have 2 sections:

• Exporting LLMs

• Exporting custom LLMs

The first one uses etLLM and the second one shows up an example with nano-gpt. We should think about the content and ordering of these export documentations.

I would think we want 3 options:

• Exporting LLM using etLLM recipes
  • Only support a handful models but have the best performance, good backend support
• Exporting LLM using optimum-executorch
  • Good integration with transformers model zoo but limited quantization and backend support
• Exporting custom LLM
  • Most flexible, but very advanced and for power users only.

Suggest a potential alternative/fix

No response