Polish the cpu mocked app to make sure the api is compatible

### 🚀 Feature Description and Motivation

We need to revisit the mocked cpu app for dev purpose.

- Let's make sure if cover most of the endpoints we like to support
- make sure the api can be well maintained if someone likes to add new endpoints
- add flexibility on the models etc.
- Have enough test coverage using openai sdk client (sync/async streaming etc)

#### Additional problems
1. vllm endpoint could be inconsistent with openai standards
2. vllm has some it's own api we need to maintain

### Use Case

for development and testing purpose

### Proposed Solution

_No response_