feat: import missing backup records from S3-compatible object storage

Summary

Add an admin-triggered workflow to rebuild missing official backup records from S3-compatible object storage.

This fixes cases where backup files already exist in the configured backup bucket/prefix, but corresponding backup_records rows are missing after migration, manual upload, old deployments, or database recovery.

Changes

• Add backup import preview API: GET /api/v1/admin/backups/import-preview.
• Add import API: POST /api/v1/admin/backups/import.
• Add admin UI to scan object storage first, then confirm import.
• Import only missing .sql.gz backup objects and skip records that already exist.
• Create imported records with status=completed, backup_type=postgres, and triggered_by=imported.
• Store file_name, s3_key, size_bytes, and fallback timestamps from object last modified time.
• Set expires_at for imported records based on the current backup retention policy.
• Keep imported historical records out of automatic count-based cleanup.
• Keep existing backup download, delete, and restore flows unchanged.
• Add pg_dump and psql to runtime Docker images and make restore fail fast with ON_ERROR_STOP=1.

Tests

• go test -tags unit ./internal/service -run 'TestBackupService_ImportMissingBackups|TestBackupService_ImportMissingBackups_BackfillsImportedExpiry|TestCleanupOldBackups_KeepsImportedRecords|TestStartRestore_Async|TestStartRestore_AsyncFailure' -v
• docker build --target frontend-builder -f Dockerfile .
• docker build -f Dockerfile .
• docker build -f deploy/Dockerfile .
• docker build -f backend/Dockerfile backend

Manual Verification

• Verified locally with Postgres, Redis, MinIO, and the app container.
• Verified on NAS with existing object-storage backups:
• Preview found object-storage backup candidates.
• Import created missing official backup_records.
• Imported records appeared in the official backup list.
• New backup restore completed successfully after adding psql to the runtime image.