Charts should specify priorityClassName by default

[d-uzlov]

Currently default values set priorityClassName empty both for controller deployment and for node daemonset.
This can lead to issues with preemption, though I haven't encountered them.

This also leads to issues during node shutdown.
K8s supports Graceful Node Shutdown. It first evicts all non-critical pods, then critical pods.
When priorityClassName is empty, controller and node pod of democratic-csi are evicted before other pods are deleted and volume unbind/detach happens.

When node is shutting down, it still holds attachments for all volumes. Pods that use ReadWriteOnce PVCs are unable to restart while node is not available.
Additionally, I found out that node hangs on shutdown if nvme devices are still connected. I feel like there is an issue with nvmet library/module here, but it wouldn't happen if nvme devices were properly unmounted.

Setting controller.priorityClassName=system-cluster-critical and node.priorityClassName=system-node-critical fixes the issue.

I checked out other csi projects: nfs-csi specifies priorityClassName, smb-csi specifies priorityClassName