MLFlow compatibility: Candidate and Experiment deletion

From #381660 (comment 2086035417)

Hi !

I'm working on a self-hosted instance, and it seems like there is no way to delete runs/experiments from the API. This affects me directly, since it means that I cannot test in a CI/CD pipeline the code that handles the interconnection between my service and your MLFlow backend without leaving trash behind. I am able to erase the artifacts (via the packages API), which does free most of the space my tests take, but not all.

I would like for delete run and delete experiment to be officially supported.

Our mlflow compatibility layer does not implement delete_run and delete_experiment, only delete_registered_model