feat: remove Azure CSI driver resize workaround - #7297

/test limit=local

@armru, here's the link to the E2E on CNPG workflow run: https://github.com/cloudnative-pg/cloudnative-pg/actions/runs/14309402140

There's a whole section of the documentation that we need to rethink.

### Expanding PVC volumes on AKS

When triggering two operations (VM +disk update) at the same time, the operator asks the pod restart for VM upgrade to complete. Azure preserves both PVC conditions at the same time, while AWS & GCP removes the Resizing one when it adds the FileSystemResizePending. This will force the operator to wait and not recreate the pod on the new VM.

diff --git a/pkg/reconciler/persistentvolumeclaim/resources.go b/pkg/reconciler/persistentvolumeclaim/resources.go
index 1ca30256f..bcaaef34a 100644
--- a/pkg/reconciler/persistentvolumeclaim/resources.go
+++ b/pkg/reconciler/persistentvolumeclaim/resources.go
@@ -65,15 +65,21 @@ external:
        return true
 }

-// isResizing returns true if PersistentVolumeClaimResizing condition is present
+// isResizing returns true if PersistentVolumeClaimResizing condition is present and not PersistentVolumeClaimFileSystemResizePending
 func isResizing(pvc corev1.PersistentVolumeClaim) bool {
+       isResizing := false
+       isFileSystemResizePending := false
+
        for _, condition := range pvc.Status.Conditions {
                if condition.Type == corev1.PersistentVolumeClaimResizing {
-                       return true
+                       isResizing = true
+               }
+               if condition.Type == corev1.PersistentVolumeClaimFileSystemResizePending {
+                       isFileSystemResizePending = true
                }
        }

-       return false
+       return (isResizing && !isFileSystemResizePending)
 } 

I agree with @nfaust-edb , this PR is not only about removing support for a feature we don't need anymore due to Azure having improved (finally) the persistent volume expansions not requiring anymore a disk to be in a "detached" state from the node to progress, but also about fixing a corner case in which if both the disk and the node are changed at the same time, we end in a state in which the new pod is not getting created because of the PVC still having the Resizing condition set to true, and this only happens on Azure (not on AWS nor on GCP) of the 3 major CSPs.

@nfaust-edb @valeriodelsarto I see your point about the proposed change, but I believe it deserves its own issue and PR for easier tracking. Could you open a dedicated issue?

Build Error! No Linked Issue found. Please link an issue or mention it in the body using #<issue_id>

Build Error! No Linked Issue found. Please link an issue or mention it in the body using #<issue_id>

Thanks @mnencia, @mbiagetti created one here #7324. I was hoping that this PR would have also fixed that issue, but needs to be confirmed by doing a test.