FEAT: Add support for Flexible Model

[shellc]

Flexible Model is a customizable inference pipeline within xinference, supporting a range of tasks including image segmentation, traditional machine learning regression, and more.

Mechanism:

Create the model object with custom definitions and initialization functions.
Use xinference.model.flexible.launchers.transformers to load transformer models.
Invoke the model via the /v1/flexible/infers endpoint.
Access the model using the SDK’s infer interface.

Examples:

from xinference.client import Client

client = Client("http://localhost:9997")

model = client.get_model("flexible-model")

result = model.infer(inputs="happy")

[qinxuye]

@yiboyasss Could you review the frontend part of this PR?

[qinxuye]

I left some comments