Multi-printer USB power: per-printer hub/port mapping

[szrudi]

What

The USB power-save feature (v1.5.x) discovers one Dymo printer's USB hub/port by matching vendor:product ID (0922:1002) and powers that port on/off via `uhubctl`. For setups with two or more DYMOs plugged into the same Pi, this breaks down — there's no way to know which port belongs to which printer, so the feature can only manage the first one it finds.

Why

• Real multi-printer deployments exist (the multi-printer UI selector has been in place since v1.4).
• Without per-printer mapping, the power-save daemon would either power off all DYMOs together (bad) or pick one arbitrarily (also bad).
• No demand reported yet, but the limitation is in the comments at `server/usb_power.py` and should be tracked here too.

Implementation sketch

• Extend the persisted USB power state (currently a single hub/port pair) to a map keyed by `PrinterInfo.id` — same id used everywhere else for printer routing.
• On discovery (`uhubctl` scan + `DeviceManager`), match each Dymo by its USB bus/address (which is what the device manager already gives us in the ID) rather than vendor:product alone. Store the resulting hub/port under that printer's id.
• Manual `/api/power/{status,on,off}` endpoints take a `printer_id` query param (default = currently-selected / first found, to keep the single-printer case unchanged).
• Auto-idle: each printer's idle timer is independent; activity to one printer doesn't keep the others awake.

Where in the code

• `server/usb_power.py` — discovery + state. The "TODO: per-printer port mapping" comment is already there as a marker.
• `server/app.py` — `/api/power/*` endpoints need an optional id arg.
• `server/power_save.py` — idle timer would become per-printer-id.
• Client: `PowerToggle` component could show one toggle per printer (or just the currently-selected one).

Out of scope

• Discovering hubs that don't support per-port power switching (ppps). Those printers simply opt out of the feature; that's already the case.
• Mixed virtual + real setups — virtual printers don't have power state.

Pin this if a multi-printer user actually runs into it.

[szrudi]

Design note: key the per-printer hub/port map by the #40 serial identity

When this is built, reuse the stable printer identity from #40 rather than inventing a parallel scheme, so settings, power, and status all key off the same thing.

• Identity vs attribute: the serial (serial:<sn>, from Key per-printer settings by serial number, not the unstable USB bus/address id #40's _printer_id) is the canonical printer identity; (hub, port) is a per-printer attribute used for uhubctl control. Don't key the printer by (hub, port).
• Discovery: uhubctl's device descriptor already carries the serial — e.g. Port 3: ... [0922:1002 Dymo DYMO LabelManager PnP 08383504012013]. Extend the existing _PORT_DEVICE_RE parse (today it captures only VID:PID) to also grab the trailing serial token, so each present Dymo maps to its (hub, port). With ≥2 Dymos, VID:PID alone matches both ports — the serial is the disambiguator.
• Persistence: replace today's single top-level hub/port in the state file with a serial → {hub, port} map — the same shape as the per-printer settings printers map (Persist per-printer label settings (tape size, colors) across visits #20). Caveat: while a port is powered off the device is absent and can't report its serial, so the persisted map is the source of truth for wake-up (same role the single cache plays today).
• Single-printer stays as-is: for one Dymo, VID:PID discovery is already unambiguous and adding serial buys nothing — this only matters once multiple printers are in play.

Related: #40 (stable serial identity), #20 (per-printer settings, same map shape), #38 (status/alias, same identity).