replay: remove GPS relative timestamps

[CarlOlsson]

Replaces #14092

Describe problem solved by this pull request
Solves original issue reported here

Describe your solution
Remove relative timestamps of GPS measurements. The benefit of this is that all GPS data that is logged will actually be published during EKF replay even if the ekf2_timestamps topic has dropouts. The downside is that the accuracy of the measurement timestamps is reduced slightly.

Describe possible alternatives
Publish the GPS data anyway if no relative timestamp is found during replay. The upside would be more accurate time stamping and the downside more data to log

Tests
Here are the relative timestamps of two GPS modules (logged in microseconds / 10)

As can be seen the difference from the topic original timestamps is on average ~1 ms. This in comparison to the ~100 ms delay of the GPS measurements themselves should have minimal effect on the replay result.

Here is the dt of the gps topic after replay, PR (top) vs master (bottom). As can be seen the dropouts are significantly reduced

And here is the selected GPS module, as can be seen the GPS timeout is no longer triggered in the replayed logfile

The PR has been tested on code similar to v1.10.1

Thanks @bkueng for the discussion!

[CarlOlsson]

On a sidenote, publishing any sensor data in replay even if no relative timestamp is found would be great to test since it might robustify replay a lot.
E.g. Here are delta velocity biases from a replayed (top) logfile vs the original

As can be seen the output is not the same

[bkueng]

Thanks! Let's get this in then and see if we can further improve things.