Create pidfile and "bpm pid" command

[jandubois]

Some drain scripts send signals directly to the process via the PID from the pidfile (because bpm doesn't have a mechanism to send e.g. a USR1 signal). This obviously only works when the drain runs inside the same container as the process itself.

This commit maintains a pid file at the /var/vcap/sys/run/bpm/$JOB/$PROCESS.pid location and also implements the bpm pid command to access it. This is an extension to the changes in #10 and should enable some of the drain scripts mentioned in cloudfoundry-incubator/kubecf#1728 to work without any local patches.

Manual testing

Start the new container-run as before (main.sh from #10):

# ./container-run --job-name main --process-name main $PWD/main.sh; echo $?
Start
Loop 1
[...]
Loop 10
End

Then run from a different terminal:

# bpm start main; sleep 3; bpm pid main; kill -quit $(bpm pid main); sleep 3; kill $(bpm pid main)
20968
# sleep 3; bpm pid main
Process is not running

And verify it has the expected effect:

[...]
End
Start
Loop 1
Loop 2
SIGQUIT
Loop 3
SIGTERM
Exiting because already received SIGQUIT

[andrew-edgar]

I have looked at most of the tests that were failing (other than the path issues) and we need to change the tests to "Expect" Pid function to called. Most of the tests have Expects for "Wait" but it will also need "Expects" for "Pid" to be called. I can add that and I know how and where to add it but none of the tests will work until the hard coding of the paths is fixed

[manno]

LGTM