improve large model inference documentation

[lxning]

📚 The doc issue

There are some issues with large model doc.

1. architecture picture(ts-lmi-internal) is broken in https://pytorch.org/serve/large_model_inference.html

2. organize core features together in https://pytorch.org/serve/large_model_inference.html

• streaming response via HTTP and GRPC

• job ticket

3. describe parameter configuration (this section is not limited to large model, it is better to have model configuration section)

• model configuration file introduction
• model config parameters definition

Suggest a potential alternative/fix

No response