Longhorn: Support automatic rebalance the replicas

[yasker]

Add a global setting to enable Automatic replica rebalance.

In the case of there is a limited or no redundancy for the volume (e.g. due to node offline temporarily), if this option is enabled, we can automatically check if a new replica can be scheduled on the new node to restore the redundancy. Notice this can have some system level impact since it will result in more network traffic in the system.

Should be implemented with Maximum Concurrent Replica Rebuilding Count option.

from: https://github.com/rancher/longhorn/issues/570

[ironsalsa]

Adding my vote here for this feature - it would be great to not need manual intervention after node failures.