resque-pool doesn't shut down manager after kill -QUIT

[wollkind]

I believe there is a bug in the handling of kill -QUIT that causes the manager to hang after all the processes are reaped. The issue is that the trap for kill -QUIT passes a zero as the flag to waitpid2, which is sensible since we do want to wait for all the child processes to end before shutting down the manager. Unfortunately this causes a hang after all the children have finished because a call waitpid2 with a flag value of zero and no living children will never return. It seems an assumption was made that this call would just return a nil pid after all children have ended but that is not the case per the Ruby docs and some experimentation. (In fact, the Ruby docs say that a waitpid2 with no children can result in a SystemError on some platforms, but this does not seem to be the case on OSX or Centos for me)

I think this code needs to be rewritten to instead wait for a known set of child pids to report that they are dead and then exit the loop. For now, kill -INT works just fine for my purposes (I'm ok with god starting up a new master before all old workers are finished)