Fix WithContainerStepTest.stop and WithContainerStepTest.death

[dwnusbaum]

I saw these flake in #256 and didn't think much of it, but then I saw them also fail in #253, and took a look. There seem to be a lot of problems with our use of ps.

The real head scratcher for me though is that the tests pass locally with httpd:2.4.12. They only fail if I use httpd:2.4.35 or newer (which I believe is due to the switch to a -slim version of Debian in docker-library/httpd#113 which removes ps from the container entirely, but it is hard to track things down because that repo does not have tags/releases). If I do update to 2.4.35 or newer then I get the exact error messages seen in CI. Is something in the CI environment forcing newer images to be used? Is that even possible?

• Make sure you are opening from a topic/feature/bugfix branch (right side) and not your main branch!
• Ensure that the pull request title represents the desired changelog entry
• Please describe what you did
• Link to relevant issues in GitHub or Jira
• Link to relevant pull requests, esp. upstream and downstream changes
• Ensure you have provided tests - that demonstrates feature works or fixes the issue

[dwnusbaum]

Let's rebuild and hope that WithContainerStepTest.stop fails this time.

[dwnusbaum]

It looks like the build might fail to record Linux test results, but WithContainerStep.test failed this time, and the error message in kill is java.io.IOException: failed to run ps. stderr: qemu: uncaught target signal 11 (Segmentation fault) - core dumped

[car-roll]

doing this whole parsing looks so painful to me 😢
Would it be easier to test if the heartbeat file still existed/was being written to?

[dwnusbaum]

I think this test is trying to check whether sh steps inside of withContainer respond correctly if the remote script just suddenly dies, so it is using internal knowledge of durable-task to kill all of the processes related to the sh step from within, which is pretty awkward. It used to be simpler before jenkinsci/durable-task-plugin#49, but it had to be updated in #121 since the pid file didn't exist anymore.

I think that if we deleted the heartbeat file here it wouldn't do anything, since the control script would just touch the file 3 seconds later, recreating it. Not 100% sure though. Maybe we could chmod it to be unwritable or something, but IDK if that would be much easier to maintain.