feat(vi/cvi): recover VirtualImage and ClusterVirtualImage from ImageLost when data returns

[danilrwx]

Description

Automatically recover VirtualImage and ClusterVirtualImage that entered the ImageLost phase (image is missing in DVCR) without re-importing the data.

The ImagePresenceHandler already moved a Ready image to ImageLost when it disappeared from DVCR. It is now symmetric and also handles the reverse transition:

• while the image is ImageLost, the handler keeps polling DVCR (RequeueAfter), so the return of the data is noticed shortly after it happens;
• once the image reappears in DVCR (for example, when the DVCR backing PVC is remounted after a temporary loss), the resource is restored to Ready — only the phase and the Ready condition are flipped, the rest of the status (Target, Size, Format, …) was never cleared;
• a VirtualImageLostRecovered / ClusterVirtualImageLostRecovered event is emitted on recovery.

Because recovery reuses the blobs already present in DVCR, no re-download happens and every source recovers, including Upload, whose data cannot be re-fetched.

The LifeCycleHandler early-returns on ImageLost as before (no import restart). Handler order (LifeCycleHandler → ImagePresenceHandler) is unchanged and the ImagePVCLost phase is not affected. Healthy Ready images are not polled — loss detection stays event-driven as before; polling is added only while a resource is lost.

Why do we need it, and what problem does it solve?

When the DVCR backing storage (its PVC) is temporarily lost and later comes back, all images served by DVCR are physically intact — only the registry was unavailable for a while. Today VI/CVI correctly move to ImageLost during the outage, but then stay stuck there even though the blobs returned, forcing users to delete and recreate the resources by hand.

This is especially painful for a mass DVCR outage affecting dozens of images at once. Re-importing is both unnecessary (the data is already there) and impossible for Upload sources. Flipping back to Ready when the data reappears restores the images automatically, with no re-download and no user action.

What is the expected result?

1. Create a VI/CVI backed by DVCR, wait for Ready.
2. Make DVCR lose the data (detach/remove the DVCR PVC, or delete the blob), wait for the resource to enter ImageLost.
3. Bring the data back (remount the DVCR PVC / restore the blob).
4. Within the recheck interval the resource returns to Ready on its own; a VirtualImageLostRecovered / ClusterVirtualImageLostRecovered event is recorded. This works for every source type, including Upload.
5. While the data is still missing, the resource stays in ImageLost and DVCR keeps being rechecked periodically.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested in the Kubernetes cluster manually.

Changelog entries

section: core
type: feature
summary: Automatically restore VirtualImage and ClusterVirtualImage from the ImageLost phase once the image reappears in DVCR, without re-importing the data.