Initially the intent for adding this was to better test
any patches for roles together before they merge, but it
has had the unintended side-effect of causing patches to
take much longer to merge (because they all get lined up
in a single queue, rather than independent queues) and
a lot more infra resources are used (because a patch
that fails at the top of the queue will result in all
subsequent patches restarting all their tests).
As discussed in the channel, we'd prefer to revert back
to the previous independent queue method of testing. It
has served us well.
Change-Id: I636b0900ea2ab001f9363027becd40085653a7f4
70ff281931