Report device health metrics for AWS leased devices

[guangy10]

🚀 The feature, motivation and pitch

Per the discussion with AWS, they don't have API to either collect or query the device health metrics we care regarding to provide stable benchmark metrics from run-to-run. So we will need to build this muscle ourselves and making it a standard process that running both prior and post to an actual on-device benchmark run.

Alternatives

AWS team suggests this is doable via test spec.

Additional context

This would be useful to root cause the metrics instability issue caused by devices regardless the device is leased from AWS public pool, private pool or even user-owned device.

RFC (Optional)

No response

cc @huydhn @kirklandsign @shoumikhin