Complete backup feature for VMI: add BackupStrategy, documentation, and UI support #1973

Parent issue: #1713

Summary

The backup API refactoring introduced BackupClass as a cluster-scoped resource and simplified BackupJob/BackupPlan by replacing strategyRef and storageRef with backupClassName. The core API work is complete, but integration into applications and user-facing components remains.

Remaining Work

1. Add BackupStrategies to application charts

• Add implemented BackupStrategy resources to the -rd (resource definition) charts for each supported application, describe as Parameters of a BackupClass strategies all that not break controller backup/restore logic (example VMI: storages, timeouts, ttl, but keep sets of resource and matches untouchable, because on this relies controller logic)
• Ensure strategies are properly configured for application types (VMs, VMDisk)

2. Documentation

• Document how to configure Velero as a backup backend
• Document how to create and configure BackupClass resources
• Provide examples for common backup scenarios

3. Dashboard UI

• Update the dashboard to support backup configuration via BackupClass
• Allow users to create and manage BackupJob and BackupPlan resources through the UI
• Allow users trigger restore (feat(dashboard): add restorejob related views #2387)

4. e2e-test

• Cover main functionality of the backup and restore VMI

5. API design issues

• Backups has VM-spectific .status.underlyingResources
• Allow user choose app-specific options in RestoreJobs (example: keep original PVCs/IPs) (added here feat(backups): restore vmi to copy in another namespace #2329)

6. Other bugs

• restore to copy is not working now (Velero DataUploaders can't resrtore to PVC with a new name, ResourceModifiers not applicable in this case, because will applied after upload) (added here feat(backups): restore vmi to copy in another namespace #2329)
• prepareForRestore is not idempotent — re-execution corrupts PVC state
• if user creates a new RestoreJob with the same name, the old -orig PVCs from a previous run might already exist, and the PVC rename logic silently no-ops. This means the Velero Restore would find the restored-but-failed PVCs instead of the originals.

Related Work

This continues the backup API refactoring that introduced:

• BackupClass (cluster-scoped resource)
• Simplified BackupJob.spec.applicationRef with default apiGroup: apps.cozystack.io