Fix MLflow tracking links: add ML Studio URLs, fix tracking URI bug

[Copilot]

AzureML serves an MLflow-compatible API server only — it has no MLflow UI. Clicking run links constructed from the tracking URI (e.g. https://westeurope.api.azureml.ms/mlflow/v2.0/.../experiments/<id>/runs/<id>) returns HTTP 404.

Changes

• 5 MLflow example notebooks (xgboost_classification_mlflow, keras_mnist_with_mlflow, xgboost_nested_runs, xgboost_service_principal, model_management): added notes clarifying the tracking URI is an API endpoint, not a UI, and added code cells that generate a working ML Studio URL after each run:

import re
tracking_uri = mlflow.get_tracking_uri()
match = re.search(
    r'subscriptions/([^/]+)/resourceGroups/([^/]+)/providers/Microsoft\.MachineLearningServices/workspaces/([^/?]+)',
    tracking_uri, re.IGNORECASE,
)
if match:
    sub, rg, ws_name = match.groups()
    print(f"https://ml.azure.com/runs/{run.info.run_id}"
          f"?wsid=/subscriptions/{sub}/resourcegroups/{rg}/workspaces/{ws_name}")

• xgboost_service_principal.ipynb: fixed silent bug where mlflow.set_tracking_uri = ws.mlflow_tracking_uri (attribute assignment, replaces the function) was used instead of mlflow.set_tracking_uri(ws.mlflow_tracking_uri) (function call).