Run `mlflow.end_run()` after ending the training process

[napulen]

It seems that running run_experiment() twice in a row will crash, because the mlflow experiment remains open. Every successful iteration of the run_experiment() function should end by concluding the experiment.