v1.11.0 — Unified command path and MQTT dispatch refactor

Changed

• Command path unified in the entity base classes. valve.py, switch.py, lawn_mower.py, automower_button.py, automower_switch.py, automower_select.py, automower_number.py, and automower_lawn_mower.py previously each duplicated the same throttle → budget-increment → try/except-auth/exception → raise-translated-error boilerplate around every API call — ~25 lines per call site, ten call sites. The shared pattern now lives in GardenaEntity._async_execute_command (Gardena side) and AutomowerEntity._async_execute_command (Automower side), with per-platform exception mapping. Call sites now read as a single line: await self._async_execute_command(client_method, *args, **kwargs). Net change: ~180 lines of duplication removed. Behaviour is unchanged — the pessimistic budget-increment-before-await guarantee from v1.10.4 is preserved.
• MQTT command handler rewritten as a dispatch table. GardenaCoordinator._async_handle_mqtt_command used a six-branch if/elif cascade with near-identical await self._client.async_send_command(…) calls. Replaced with _MQTT_DISPATCH: dict[str, tuple[control_type, command, needs_duration]] and a single dispatch call — adding a new MQTT action is now a single-line change. The redundant _MQTT_ACTIONS set is removed.
• automower_lawn_mower.py command handler also gets the same treatment: the five-branch if/elif inside the try/except is replaced with _COMMAND_METHODS: dict[str, str] and a getattr lookup against the client.

Developer Notes

Pure refactor release. No functional or UI-visible change; all existing tests pass unchanged as the behavioural contract (throttle first, count second, dispatch third, on-success state update) is identical to v1.10.4.

Net diff: +187 / −306 (−119 lines). All 725 tests pass.