Better control of re-train when editing custom monitors

If a user is making a change to a custom monitor that might significantly alert its results, the user is now prompted if they want to retrain their anomaly detection models or not. Previously, Monte Carlo would automatically begin a retrain, and there were some cases where this was not desirable (e.g. a small change in the filter).

Users are only prompted with this modal if they are making a significant change to the SQL, the schedule, segmentation, editing variables, etc. Editing a field like notes, description, or audience will not prompt this modal.