Comment 2 for bug 1273912

Just making some notes for anyone who wants to do this:

 - need to preserve the existing command (test enumeration is needed)
 - shouldn't depend on re-invoking - we may want a command interface in future for online scheduling,so this should be explicitly handled in testr
 - dry-run is rougly equivalent to --list-tests, in that that also enumerates and doesn't run, so --list-tests --parallel could perhaps be a way forward
 - another way would be to enumerate and schedule, and output via subunit with worker tags