Possible race condition in child process cleanup logic

[borsothy]

If the SIGINT trap here happens to start between these two lines, the workers array possibly contains process IDs belonging to already stopped processes. This results in an Errno::ESRCH exception.

So we're right here:
dead_guys = Process.waitall
and then the main process receives an INT signal, delegating that to the child processes:
Signal.trap("INT") { workers.each { |w| Process.kill("INT", w) } }
but the dead PIDs are removed after the very first line:
dead_guys.each { |pid, status| workers.delete(pid); puts pid.to_s + " exited" }

[borsothy]

Proposing fix in #120