Improve signal-driven notification system

[kako-nawao]

We have a few signal-driven (aka event-driven) notifications, which are static (built into the code). We should probably extend this to allow customizing listeners, setting up conditions and whatnot (alert me if activity X reaches 10000 approved hours).

The concept would be:

• Something "important" happened, we signal it ("timesheet.approved")
• A generic listener captures it and delegates it to the CustomListenerDispatcher
• The dispatcher selects all listeners for that event (Listener A and Listener B)
• Listener A checks that the pre-condition to execute its action is not met, it stops
• Listener B checks that the pre-condition to execute its action is met, it executes its callback(s?)

I implemented something along those lines once, but it was using a custom event subsystem (not django signals) and storing them in MongoDB. This could be done without it turning into a nightmare if conditions are simple, something like {actor_x.attribute_x} {relation} {value | actor_y.attribute_y}. We'd also need to validate listeners somehow, to ensure that they won't create endless loops.

[omelendrez]

I like that. It's very marketable. The project manager can receive an alert if some activity reached a specific percentile of consumed hours. Also, from the equipment point of view, Workshop would like to be informed if an equipment had been used for specific number of hours so it can go for maintenance.