The upgrade process should not depend on storage class for detecting shared PX volumes

[harsh-px]

BUG REPORT:

waitTillPXSharedVolumesDetached() https://github.com/portworx/talisman/blob/master/pkg/cluster/px/px.go#L435 looks for all storage classes that have shared: true and uses that to find PX shared volumes. Instead it should talk directly to Portworx and enumerate PX volumes that are shared.

What happened:
waitTillPXSharedVolumesDetached failed to detect 2 PX shared volumes that did not have shared: true in the storage class. Hence it went ahead with the upgrade process without detaching these.

How to reproduce it (as minimally and precisely as possible):

1. Install Portworx 1.3.1.4
2. Create a non-shared repl 3 portworx volume. Later on update this volume to be a shared volume. (Alternately, create a shared volume though a storage class and then remove shared: true from the Storage class).
3. Have an application pod attach and mount the volume
4. Use talisman to upgrade to 1.3.4

At step 4, talisman will fail to wait for this volume to be detached and will end up triggering the upgrade.