ack behavior improvements

I have clients who rightly expect that if they ACK something at warning, they will get a new alert if the issue worsens to error or critical level.  I have to explain to them that LM doesn't work this way, because once something is ACK'ed, it is ACK'ed forever (until the alert clears), at least it certainly seems to be the case and there is no documentation to the contrary.  Having used other tools that do support more sophisticated methods, I recommend adding the following features:

* option to reset an ack when state becomes worse (this can admittedly open a can of worms when states transition rapidly, so perhaps this should still require a delay timer)
* option to manually clear an ack that might have been set by mistake
* option to set a maximum time for ack's to remain active (hierarchical/inherited timer value, definition set in datapoint when desired).  this is because some issues are so important you don't want to let them just become ignored indefinitely in error.

Some of these can be managed via API-based enhancements, but it would be preferred to have direct controls.

Thanks,
Mark