Added pitot tube icing detection

[RomanBapst]

Solved Problem

When ice starts to build up at the front of the pitot tube, usually this results in a scale error of the airspeed reading which slowly increases over a certain amount of time. The current checks are not well suited to detect this kind of failure, as the detection of a failure via the airspeed innovations checks is usually not fast enough. Airframes which suffer from structural damage caused by limit cycling due to inaccurate airspeed readings used for control surface scaling, rely on a rapid detection of this failure case.

The default empirical thresholds proposed in this PR stem from real world testing on a platform which suffers from the aforementioned failure case. Since real icing conditions are hard to reproduce, this feature has been validated by injecting scale factor errors which increase over time. The failure has been detected consistently with ramp times of 2, 5, 30, 60 and 120s from a scale factor of 1 to 0.

Results This shows the test run we did on a vehicle by injection airspeed scale errors (ramping up the scale error using different ramp times). Note that initially the scale error is ramped up really quickly and the innovation checks would easily trigger.

This figure shows the results when the scale error is ramped up slowly. Note how the innovation check does not really help in this case but the derivative check still triggers.

Changelog Entry

Added first principle based airspeed failure check to detect growing scale errors caused by icing. 
New parameter: ASPD_FP_T_WINDOW