feat: expose Q10 current cleaning room

[hCoureau]

Q10 firmware does not provide a validated active-room field. This exposes map.current_room by conservatively matching the latest live robot position to the segmented map grid.

Behavior

• Return the existing Q10Room model or None, with a defensive copy of a resolved room.
• Use the exact segmented cell when available. A bounded nearby-cell search handles unsegmented positions; ambiguous results remain None.
• Prefer usable header calibration, with path-fit calibration as a fallback for older or incomplete packets.
• Require an active or paused cleaning state and a fresh live trace. Inactive/docked updates clear freshness so a prior session is not reused.
• Preserve custom room labels and redact raw room names in diagnostics.

Session validity stays in the trait. Spatial resolution is a pure function of typed map/trace models.

Dependencies

Depends on #938 and #937. #936, #908, and #965 are merged. Upstream main through b3a98b9 (7.12.0) is integrated.

Allen’s follow-up feedback on merged #936 is addressed by separate clean-record and saved-map decoder tests, with no conditional assertions.

Validation and limits

Validation of 33b5af5:

• uv run pytest -q on Python 3.13: 1,497 passed, 53 xfailed, 92 snapshots passed.
• All pre-commit hooks passed, including Ruff, mypy, codespell, and structured-file checks.
• uv build passed for both wheel and sdist.
• The affected cumulative protocol suite passed on Python 3.11 and 3.14: 96 passed, 10 snapshots passed on each version. The full cumulative suite passed on both versions before this test-only follow-up and merge.

No physical-device run was performed for this revision.

Existing tests cover room resolution, ambiguity, calibration, session freshness, custom labels, and diagnostic redaction. Earlier physical Q10 S5/S5+ testing observed a custom room label during a single-room clean and clearing during return/docking. A room-to-room transition was not covered by that reported run.

This is inferred room occupancy during cleaning, not an authoritative firmware report of the selected cleaning segment. Missing or unusable map/trace data returns None. Archive requests inherit the bounded lifecycle from #937, including its clean-history ambiguity barrier and same-map freshness limitation.

No private map captures, room labels, or account data are included. Related: #767.

[lboue]

Tested against a real device: Roborock Q10 S5+ (ss07 firmware family), on commit cd2621d (this PR's isolated commit, rebased onto its stacked dependencies).

Watched device.b01_q10_properties.map.current_room live via add_update_listener while running a targeted clean:

{"time": "13:16:10", "state": null,            "room_id": null, "room_name": null,        "position": null}
{"time": "13:16:11", "state": "SWEEP_AND_MOP",  "room_id": null, "room_name": null,        "position": {"x": -976, "y": -888}}
{"time": "13:16:11", "state": "SWEEP_AND_MOP",  "room_id": 3,    "room_name": "Salle d’eau","position": {"x": -976, "y": -888}}
...
{"time": "13:18:26", "state": "RETURNING_HOME", "room_id": null, "room_name": null,        "position": {"x": -412, "y": -508}}
{"time": "13:20:03", "state": "EMPTYING_THE_BIN","room_id": null,"room_name": null,        "position": null}

Observations:

• current_room correctly reflects a custom room label (Salle d’eau, French for "bathroom") rather than a raw firmware rr_... default - matches the "preserve custom room labels" goal.
• It's briefly None right after the position/state first arrive (before the first segmented-grid match resolves), then locks onto the right room.
• It correctly clears back to None as soon as the state moves to RETURNING_HOME / EMPTYING_THE_BIN, confirming stale session state doesn't leak into docking/post-clean states.

This was a single-room spot clean, so I didn't get to observe a room-to-room transition in this pass.