Improve Semantics of Post-Build Tasks

[stevekuznetsov]

Today, we run a number of post-build publishers. As only one org.jenkinsci.plugins.postbuildscript.PostBuildScript can be run as a publisher, we end up with a block of buildSteps of type hudson.tasks.Shell. This means that if we have a post-build flow like -- generate artifacts, retrieve artifacts, fetch systemd journals, deprovision cloud resources -- and one of the first hudson.tasks.Shell fails, the rest will not run. Although the larger org.jenkinsci.plugins.postbuildscript.PostBuildScript is set to run regardless of whether the job failed or not, the linear flow of hudson.tasks.Shell will exit early on any individual failure. We can try to address this by adding || true to our actions in these steps but in reality we just need a way to parameterize a named_shell_action so we don't always add set -o errexit. This way, failures will be silently ignored and all post-build tasks will run.

/cc @soltysh

[soltysh]

I need to familiarize myself with how this all machinery is working, but I'll try to hack something up.