Add support for querying Unity Catalog model services

[billiam-wang]

What

Adds a ModelServicesExt mixin exposing w.model_services.get_open_ai_client() and get_langchain_chat_open_ai_client(), so users can query model services (model APIs) through Unity Gateway. A model service is a Unity Catalog securable that represents a governed LLM endpoint. The returned client targets {host}/ai-gateway/mlflow/v1 with refreshed Databricks bearer auth; the model service's Unity Catalog three-level name (e.g. system.ai.claude-sonnet-4-5) is passed as model. This mirrors the existing ServingEndpointsExt query surface.

Usage

from databricks.sdk import WorkspaceClient

w = WorkspaceClient()
client = w.model_services.get_open_ai_client()
resp = client.chat.completions.create(
    model="system.ai.claude-sonnet-4-5",
    messages=[{"role": "user", "content": "Hello!"}],
    max_tokens=256,
)
print(resp.choices[0].message.content)

Note for maintainers

w.model_services has no backing generated API, so the wiring in the generated databricks/sdk/__init__.py was added by hand. It likely needs to move into the SDK generator config to survive regeneration.

Testing

Added tests/test_model_services_mixin.py mirroring test_open_ai_mixin.py (base URL, reserved-param guard, bearer-auth wiring). Verified locally that the client resolves to /ai-gateway/mlflow/v1 and the guard raises.

This pull request and its description were written by Isaac.