feat: add dust collection settings to the B01 Q7 api

[akj]

What

Adds two setters to Q7PropertiesApi for the auto-empty dock, following the existing prop.set setter conventions (set_child_lock, set_volume):

• set_dust_collection(enabled) — writes dust_auto_state
• set_dust_collection_frequency(cleans) — writes dust_frequency (1 = empty after every clean); validates >= 1

Plus parametrized tests in tests/devices/traits/b01/q7/test_init.py matching the existing setter tests.

Verified on real hardware

Tested against a Roborock Q7 M5+ (roborock.vacuum.sc05, fw 03.01.80, US region):

• dust_auto_state round-trips 1 → 0 → 1 via set_dust_collection, confirmed by GET_PROP read-back after each write.
• dust_frequency round-trips 1 → 2 → 1 via set_dust_collection_frequency, same method.

Deliberately not included: manual "empty now"

Writing dust_action via prop.set is rejected by the device with code 1 (B01 command failed with code 1), so dust_action appears to be read-only state. The product schema for sc05 exposes write-only DPS 201 start_clean_task, 202 start_back_dock_task, 203 start_dock_task — a manual empty trigger most likely goes through 203 start_dock_task, which the Q7 channel doesn't support sending yet. Happy to test candidate encodings against my device in a follow-up if someone knows the expected payload shape.

Side observation from the same session, can file separately: this healthy, docked device steadily reports fault: "fault_2103" (unmapped placeholder in B01Fault, adjacent to 2102 cleaning_complete) — the 2103 reading appeared right after a completed clean while charging, so it looks like another benign 2100-series notification rather than a real fault.

[akj]

Live-hardware follow-up from the same Q7 M5+: I monitored a full app-started clean (102 min, 59 m²) end-to-end over MQTT, polling prop.get at 3 s cadence through docking, and caught the auto-empty cycle:

status=DOCKING station_act=0 dust_action=0 fault=2102 # clean complete, returning
status=CHARGING station_act=3 dust_action=1 # dock emptying (~30 s)
status=CHARGING station_act=0 dust_action=0 fault=2103 # finished, docked idle

• station_act=3 + dust_action=1 are the read-only "dust collection in progress" indicators — consistent with dust_action rejecting writes (code 1). The clean record afterwards reports record_dust_num: 1.
• The dock triggers the cycle internally on docking after a completed clean; no cloud-visible command precedes it, and a stopped clean does not trigger it (tested).
• 2102 (cleaning_complete) fired exactly at completion/return; 2103 fired once the task finished on the dock and persists while docked-idle. Both are task notifications, not faults — fix: name B01 Q7 code 2103 and document dock dust-collection state fields #921 names 2103 and documents the dock fields.

Happy to run more captures on this hardware if other codes or values need confirming.