Snapshot, restore, and delete

Snapshots are managed from an instance's detail panel. On the powerpool btrfs pool, copy-on-write makes them effectively instant, and each operation streams live progress into the panel.

• Create snapshot — name a point-in-time snapshot of the instance.
• Restore snapshot — roll the instance back to a snapshot. This is confirmed before it runs, because it reverts the current root state to the snapshot.
• Delete snapshot — remove a snapshot. Also confirmed.

Each action is gated on its own permission (snapshot.create, snapshot.restore, snapshot.delete).

Restore reverts the instance's root only. Data on attached persistent volumes lives outside the rollback boundary and survives a restore — the same state-boundary principle the deploy lifecycle is built on.