MLflow connector: import versions from MLflow Model Registry

[Fredbcx]

Context

Many teams already use MLflow. MLineage should be able to read an existing
MLflow Model Registry and import its versions as a LineageGraph, so users
don't have to start from scratch.

Task

Implement mlineage/connectors/mlflow_connector.py:

from mlineage.connectors.mlflow_connector import MLflowConnector

connector = MLflowConnector(tracking_uri="http://localhost:5000")
tracker = connector.import_model("fraud-detector")
# → returns a Tracker with versions populated from MLflow

Notes

• Optional dependency: pip install mlineage[mlflow] (already in pyproject.toml)
• MLflow versions don't have parent relationships — import them as a linear chain sorted by version number
• Tests can use mlflow's built-in file-based backend (no server needed)