OmniSensor.fetch_state() missing error handling

[StevenEmelander]

Bug

OmniSensor.fetch_state() overrides Interface.fetch_state() to use its own get_level() method with custom formula-based conversion (as noted in the existing comment: "We are explicitly not calling the parent methods in fetch_state and handle_object_status, as we don't want SensorInterface to handle the state").

However, the override does not include the error handling that Interface.fetch_state() provides. The base class catches CommandError and ConversionError per property getter (base.py lines 217-225):

for prop, getter in self._property_getters.items():
    try:
        fetched_properties[prop] = await getter(self)
    except (CommandError, ConversionError) as ex:
        logger.warning(...)
        continue

Since OmniSensor.fetch_state() bypasses this and calls get_level() directly without a try/except, any CommandError (e.g., NotInitializedError for sensors that aren't ready) or ConversionError propagates up unhandled. This can cause Controller.fetch_state() to fail entirely, preventing the integration from setting up.

Expected behavior

OmniSensor.fetch_state() should handle errors gracefully, consistent with Interface.fetch_state() — log a warning and continue rather than raising.

Proposed fix

Add the same try/except pattern to the OmniSensor.fetch_state() override, catching CommandError and ConversionError with a warning log. PR incoming.

[StevenEmelander]

The proposed fix in #357 is a bandaid — catching CommandError/ConversionError prevents the crash, but doesn’t address why get_level() is throwing in the first place.
This is likely related to legacy firmware (2.x) returning a non-standard response format for invoke() calls, which may result in a malformed or empty result making it through to OmniSensor.fetch_state(). I’ll follow up with a more targeted fix once I can capture the raw firmware responses and confirm the root cause.