This helps a lot with the layout of the pipelines: since bootstrapping and
each distribution pipeline are now stand-alone pipelines, we don't need to
worry about includes or naming every single stage we may need etc.

Our main pipeline just gets down to bootstrapping, then each distro
pipeline, all of which run parallel.

The distro pipelines have to wait for bootstrapping to be complete so our
pipeline is less efficient than before where non-qemu jobs could start as
soon as their bootstrapping image completed. But the pipelines are less
complicated to maintain and generated, so let's take that hit.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>