Poll Agent status when starting

[elijahandrews]

This affects the Debian and RPM Agent.

Right now when starting the Agent we start supervisor, sleep a fixed period of time (4 seconds on Debian and 7 on Red Hat), then run a check to see if supervisor has started all of the Agent subprocesses successfully. We run into issues when it takes a user's system more than 4/7 seconds to correctly start the Agent, as the init.d script will report that the Agent failed to start when it will have correctly started a few seconds later.

Instead of waiting a fixed period of time to check for running subprocesses, we should have a script that runs the check around once per second for some longer period of time. If it detects that the Agent has started the init.d script will report that the Agent has started successfully. If after the fixed period of time it does not detect that the subprocesses are running the init.d script will report a failure. This will greatly reduce the possibility of the Agent subprocess check running prematurely, and should shave a few seconds off of the Agent's startup time on most machines.

[elijahandrews]

6403c68 addresses this issue for Debian. @remh @LotharSee does this look suitable?

[LotharSee]

Looks good to me! This behavior is way better.
I didn't test it but it looks okay.