Node CPU usage, MetalK8s UI may display negative values

[thomasdanan]

Component:

'ui', 'upgrade', 'prometheus'

What happened:

During an upgrade phase, we may see some CPU usage negative value in the Metalk8s UI.

It is maybe related to the fact that some nodes being cordon/uncordon or some pods being upgraded and restarted.
Nevertheless, it shows there is something wrong in the Prometheus expression we use to collect Node CPU usage data.

What was expected:

Node CPU usage should never display negative values

Steps to reproduce

Resolution proposal (optional):