[Bug] Automatic re-calibration could converge faster

[laurensvalk]

There is always a tiny amount of expected drift after initial calibration, such as 0.01 deg/s.

The gyro re-calibrates continuously whenever it is stationary, at a rate of approximately 5% per second of being stationary.

In some cases, this appears to be not fast enough to "catch up" with the changing value.

It is worth exploring if setting the decay rate to 10% or 20% might be better.

[DrTom]

#1687 is related to this one

If the calibration would weight recent readings (much) higher than older ones we could expect that some effects of the drift caused by thermal changes is mitigated.