If a single sensor is fitted, the calculation is not performed and zero values are published.
If dual IMU's are fitted, the vector length difference between the primary IMU and the second sensor is output for the angular rates and accelerations. The vector difference is low pass filtered before the length is taken.
If three IMU's are fitted, the vector length is calculated for both alternative sensors and and the maximum values output.
Fourth and subsequent IMU's are ignored.

Use value obtained from tuning on replay logs

Conflicts:
	src/modules/ekf2/ekf2_params.c

This fixes #5855.

This reverts commit 0ca397ab.

This reverts commit c24046ca.

Fixes the case where an ack got lost and the FMU resends the data, but
the client does not resend the ack.

- during large transients in pitch demand the pitch integrator value
was shifted such that the final demanded pitch did not violate given
limits. Since this strategy can cause large knock-backs of the pitch
integrator we remove this logic completely. We already have logic in place
which reduces the integrator at the pitch time constant in case the
pitch limits are exceeded so we don't need to limit it further. This
has the advantage that spikes in the specific energy balance error
signal does not lead to integrator knock-back.

Signed-off-by: Roman <bapstroman@gmail.com>

Constrain the specific energy balance integrator input to prevent increasing saturation of pitch demand.
Decay the specific energy balance integrator state if the pitch demand is saturated to reduce saturation to zero and do so at the same tome constant as the control loop
Relax the clipping threshold on the specific energy balance integrator to allow the input constraint and decay functions to do more of the work
Improve variable naming and commenting

- when limiting the pitch integrator input the value was related to
a quantity with different units (specific energy error rate vs delta pitch)
- once the unconstrained pitch demand is larger / smaller than the max/min
allowed pitch angle the integrator input should only be allowed to drag
the integrator into the direction leading to less pitch demand violation

Signed-off-by: Roman <bapstroman@gmail.com>

- if the specific energy balance correction term produced a demanded
pitch value which exceeded the aircraft pitch limits then the pitch
integrator was shifted such that the pitch demand violation was prevented.
However, this meant that the exceeding pitch was just unloaded into the
integrator and caused unexpected behavior of the pitch loop.
In an underspeed condition e.g. this has lead to the plane pulling up it's
nose very quickly shorty after the underspeed condition kicked in.

Signed-off-by: Roman <bapstroman@gmail.com>

- the method in TECS for detecting an underspeed condition was run after
the method which calculated the airspeed demand. As a result the specific
engergy balance error signal showed a spike when TECS detected an underspeed
condition.

Signed-off-by: Roman <bapstroman@gmail.com>

In most cases, really only 1 element is needed. The dynamic allocation
handles cases where more are necessary. This is all done within a locked
state, so no races can occur.

Frees roughly 2.3KB RAM.

We expect a short round-trip time, so lowering the retry timeout will
increase throughput on links with high drop rate.

Avoid sending data before we have an ack from the logger.