Update registrar to use start_watching_system() API

[anevilyak]

Currently the registrar does a polling sanity check in order to detect teams which have gone away without unregistering properly (i.e. in the case where a team is killed/crashes rather than quitting normally). This could be eliminated by instead using the new start_watching_system() call to monitor for team creation/destruction.

[anevilyak]

If no one beats me to it, I'll try to deal with this over the weekend sometime, just wanted to file the ticket so it's not forgotten.

[waddlesplash]

That's a long weekend... :)

I might take a stab at this. Anything else that might have changed in the past 8 years that I'd need to know?

[tqh]

I'll see if I can take a stab at this, I was checking what is using cpu at idle and 2% seems to be registrar timer_thread, and that seems to be this. Changing it from 1s to 5s takes that down to 0.1% cpu. Checked with top.

Some reference for how to do it (from waddlesplash): ​http://xref.plausible.coop/source/xref/haiku/headers/private/system/system_info.h#56 ​http://xref.plausible.coop/source/xref/haiku/src/servers/launch/LaunchDaemon.cpp#488

[waddlesplash]

Fix merged in hrev52796.

[nielx]

Assign tickets with status=closed and resolution=fixed within the R1/beta2 development window to the R1/beta2 Milestone