Walk all instrument keywords, make sure there are alerts for all which need them

Make sure there are alertsActor limits, etc. for all values which matter.

There is a design issue here. We currently send all alerts, and thus all keywords which can have alerts, off to STS. And STS only knows about triggering on whether the associated text field is "OK" vs not "OK". This is fine for major alarms (detector temps, instrument power, etc). But we are proposing adding many dozens of keywords, and will generate many non-critical warnings. Now that have used alerts and STS for a while, are we still OK with this simple alerts-to-STS design? Can we improve it?

There is an old ticket wondering whether STS could easily distinguish more finely – I think I asked whether we could use regular expressions instead of just a string match. Basically, whether we could match using "OK" or "^WARNING" or "^CRITICAL", and keep the rest of any non-OK text field for humans.