During Kubernetes upgrade, if the driver deployer restarts the csi deployments when the Kubernetes version changes, Kubernetes might not be able to detach the Longhorn volumes thus blocking the drain flow.

Suggesting to not restart the csi deployments when the Kubernetes version changes.

This is the revert of #426. We no longer support the Kubernetes versions in that ticket. Also, when the Kubernetes version changes such that it uses a new CSI spec, most likely that we will deploy new CSI deployment images so this will force the csi deployments' restarts anyways.