polling: reset bad hosts on *manual* poll #7029
Description
If there are no good hosts left for a platform and a task is manually triggered, the bad hosts are reset.
I.E, if the user manually triggered a task, we should attempt a submission, even if we know it is likely to fail.
This is designed to allow users to quickly "unstick" a workflow in the event of resolved platform issues.
We should extend this to manual poll.
Note, we cannot extend this to automated actions as the resetting of bad hosts by one command can interact with another remote operation running in parallel, leading to an infinite loop. Manual operations, however, are safe (we already do this for manual trigger).
Takes the sting out of #7001, the remainder becomes a low priority.