FEAT: Add embedding models support

[aresnow1]

Add embedding models support.

• Base support for model load and encode.
• Refactor some APIs to support launching and serving embedding models.
• Add top 5 or top 10 models in hf leadboard.

[aresnow1]

Sample code:

In [7]: client = Client("http://localhost:9997")

In [8]: uid = client.launch_model(model_name="gte-base", model_type="embedding")

In [9]: model = client.get_model(model_uid=uid)

In [10]: model.create_embedding("write a poem.")
Out[10]:
{'object': 'list',
 'model': '3ef99480-496f-11ee-9009-c2c8e4cad3f6',
 'data': [{'index': 0,
   'object': 'embedding',
   'embedding': [-0.003699747147038579,
    0.019329899922013283,
    ...,
    -0.05098249390721321,
    -0.001861078548245132]}],
 'usage': {'prompt_tokens': 37, 'total_tokens': 37}}