Alert Severity based on time in alert.

This goes along with other requests previously to enable detection of thresholds over time (it may be possible to do something like this with the API, but writing dataources that use the API is a bit cumbersome).  Some resources show patterns where they go in and out of alert, resetting the counter, but if calculated over time you would clearly see a problem.  The upcoming alert capability for anomaly detection may help here, but no idea when it will be available or what features it will have.  Another related issue is events, which need correlation to know if they are actually ongoing, but it is not possible within LM currently.