Fix PID derivative kickoff and reset

[matejkarasek]

This PR fixes some unwanted consequences of #1337

Since the D term now uses the difference of current and previous measurements (instead of errors) this introduced two issues:

1. for yaw, flying near 180deg yaw results in D spikes (measurements jumping between +180 and -180)

2. the pid internal measurement state is being reset to zero, which is not true for x, y, z, and attitude angles. This resulted in a D spike at the beginning of every flight

Both of these issues get especially noticable when LP filters are being used in the PID, as the spikes will propagate in time.

For 1), I removed the pidSetError functionality, as it was only used to deal with yaw discontinuity around +180/-180.
Instead, an additional parameter in pidUpdate is now used to enable extra code inside the pid that deals with the discontinuity.

For 2), the reset now takes into account current state values (x, y, z, roll, pitch, yaw).

TODO: The reset in src/modules/src/crtp_commander_rpyt.c is not yet fixed, as I did not find an elegant way to read out the x,y,z states.

Still, even without the last TODO item, this should fix the majority of use cases...

[gemenerik]

I haven't observed the spikes in practice, though they seem logically possible. Thanks for your patience and for the PR.