problem: detect/mitigate repurposed devices

[karenetheridge]

problem: sometimes a device is added to a build and rack and it has been repurposed from somewhere else -- therefore its history in conch can be misleading, e.g. it has a history of successful reports against its prior sku, still retains memory of its old set of disks etc.

We should detect this in conch and clear out the irrelevant information, and also (when possible) flag that its hardware hasn't been properly reset by the integrators and needs additional attention before it can be preflighted.

From the conch side, my initial thinking is:
If a device is assigned to a rack layout (via the POST /rack/:id/assignment) interface, and the device already exists in conch:

• check if its hardware sku matches what the layout wants to be there. If it doesn't match, forcibly reset the sku
• move its phase back to 'integration' (currently the earliest phase)
• clear the 'validated' timestamp on the device
• clear all device_setting entries (which store the reboot count, memory status, firmware status and many other values that livesys stores for tracking internal state)
• clear out all the associated device_disk entries
• reset the device health to 'unknown' (or perhaps 'error', if we can detect that there are other issues with the device TBD -- including the device phase not already in 'decommissioned').