Fix gazebo-classic_iris_opt_flow not arming

[GuillaumeLaine]

Solved Problem

Enables the gazebo-classic_iris_optical_flow model to arm, which was currently impossible.
Previously make px4_sitl gazebo-classic_iris_opt_flow with commander arm would fail. Caused by the SITL and PX4Flow returning a 0 integration timespan before takeoff, which never summed to the desired publishing time delay and disabled publishing to vehicle_optical_flow. This caused preflight checks to fail.

Solution

Publishing frequency check now compares sensor time with last publication time instead of summing integration timespan

Changelog Entry

For release notes:

Bugfix: Fix a bug that prevented gazebo-classic_iris_optical_flow model from arming.

Alternatives

We could also alter the PX4Flow plugin to publish a sensor_optical_flow message with more meaningful/non-zero timespan.

[bresch]

This might affect this fix: efbed3b
Because the integration timespan might be slightly different than the publishing rate.
@dagar is that acceptable? Or should we rather force a minimum publication rate?

[dagar]

So arguably the real problem here is the hack that was used historically to allow us to get a valid position estimate while on the ground with optical flow (and no current valid optical flow data).

Let's see if we can find an acceptable short term solution that doesn't backtrack on #22086. I think we could just handle this case a bit more explicitly in ekf2, but first we'll need to propagate something from sensors/vehicle_optical_flow.

[GuillaumeLaine]

@bresch @dagar Just a ping. Should we close this, review it, has it been resolved elsewhere?

[GuillaumeLaine]

Closing due to inactivity