feat: Multiple inference engines for nitro and openai

[hiro-v]

• Refactor nitro inference extension
• OAI compatible inference extension
• fs.exists() abstraction
• Route model init, model request
• Refactor calling direct function to Even Based system

Ref:

• 810 docs add modeljson and revamp models specs page #816

[freelerobot]

btw @vuonghoainam can just title issues feat: ... to trigger ci. WIP is implied since your PR is a draft

[github-actions]

Preview URL: https://48b9ddec.jan-qn8.pages.dev

[hiro-v]

How to use OpenAI and Nitro backed models
===> Once the app booted up, there are engines/*.json files

OAI models

1. Update base_url and api_key. By default this is OpenAI platform API but you can change it to any OAI compatible API out there (Azure, vLLM OAI entrypoint, etc)
   
2. Go the Hub and search for OpenAI (I added 3 models there), click Use
   
3. Create a thread and start to use
   

GGUF models (backed by engine=nitro

1. Find any GGUF model in the hub
   
2. Create a new thread or use in existing thread with GGUF model, then just use it