Configure the drain-skip labels on the side of network-operator for the driver upgrade

[almaslennikov]

Currently the label for pods that allows them to be skipped during drain, is configured in the common driver upgrade library.
This approach is unclear since it's not part of the explicit public API.
We can configure our own such label by including one more expression into the podSelector in the drain spec. This logic for handling drain skip should also be removed from the https://github.com/NVIDIA/k8s-operator-libs

[nadavrub]

Q: "We can configure our own such label" - does this mean there would B redundancy w/ the CommonUpgrade lib, or should there B a corresponding fix to remove the logic from that lib?
A: There will be redundancy. I think the best course of action would be to apply the change to net-op, notify the gpu op team and drop when/if they apply the same change to their code

[adrianchiris]

support was merged in k8s-operator-libs in: NVIDIA/k8s-operator-libs@f698677

update to network operator will be done as part of PR #685