add a --force option to testflinger cancel to forcibly kill a job

So I ran into an issue where I needed to kill a job mid-provision because the node in question didn't actually PXE, instead it just booted the previously installed OS.

Turns out this kinda messed testflinger up a bit because it was still trying to do the provision from that job, leaving my new job stuck in a "Waiting" state.

I understand that killing a job mid-provision could leave the node in a bad state, but it would be helpful, for me at least, to be able to forcibly kill at any time and have TF completely kill the job, including halting any provisioning in progress (maybe with a big fat warning like "You realize this could turn your device into a brick, are you sure you want to proceed?")

Anyway, just a thought that occurred to me. It would be OK on servers to do this, but I could see where doing this to a RPi or other embedded system could make the system into a pricey paperweight.