Report feature_importance speedups
Two changes/additions:
- Speed up idq-report, specifically in the feature_importance section. If
--skip-timeseries
is not set, we only look at the feature importance plots for the span in which the models were actually evaluated for rather than the entire report span specified. This can speed up the runs considerably. - Added an
elapsed_time()
func in utils.py for convenience in timing/debugging. In order to use this, add the@elapsed_time
decorator on top of the function/method you want to time.