Experiment Tracking for Kubeflow SDK

### What you would like to be added?

As part of the [centralized experiment tracking initiative](https://github.com/kubeflow/community/pull/892), we need to provide experiment tracking APIs to the Kubeflow SDK that enable AI Practitioners to log experiments, runs, metrics, parameters and artifacts directly from Python code.

This would allow users to track experiments from Jupyter notebooks while storing metadata in Kubeflow's centralized [Model Registry backend](https://github.com/kubeflow/model-registry/issues/1224).

### Why is this needed?

Providing seamless experiment tracking experience is critical for AI Practitioners who need to log and compare experiments across different stages of the ML lifecycle. Supporting MLflow compatibility makes this functionality accessible to the broader ML community who are already familiar with MLflow APIs, reducing adoption barriers for Kubeflow.

### Love this feature?

Give it a 👍 We prioritize the features with most 👍

Start by reviewing the centralized experiment tracking initiative and the Model Registry backend issue linked in the description. Define the SDK APIs needed from Python and Jupyter notebooks for experiments, runs, metrics, parameters, and artifacts, with MLflow compatibility; done means these workflows can store metadata in the centralized backend.

Written by the indexing model from the issue text.