gradio test draft

[april-yyt]

Description of changes:

• wrapped the runtime and model initialization in incremental decoding and spec infer into a Langchain formatted class
• integrated with gradio and chainlit for chatbot interfaces, following tutorials:

1. gradio: https://www.gradio.app/guides/creating-a-chatbot-fast
2. chainlit: https://docs.chainlit.io/integrations/langchain

Related Issues:

1. gradio: reponse generate functionality not working when model initialization is called outside of the "predict" function
   for example:
   [would not work]
   model = model_init()
   def predict(message):
   return model.generate(message)
   [would work] but is not what we want since the model should only be initialized once
   def predict(message):
   model = model_init()
   return model.generate(message)
2. chainlit: integration undergoing debugging

Linked Issues:

• Issue Legion error during integrating Gradio #1203

Issues closed by this PR:

• Closes #

---

This change is 

[goliaro]

@april-yyt I ran into the same issues when debugging on my machine. Perhaps we can look at the Ray Serve integration (https://docs.ray.io/en/latest/serve/tutorials/gradio-integration.html), they might have had to solve similar issues

[jiazhihao]

@april-yyt Do we still need this PR? Otherwise, I will close it.

[april-yyt]

@april-yyt Do we still need this PR? Otherwise, I will close it.

We don't need it anymore and feel free to close it, thanks!