EKF2 skipping magnetometer axes upon other axis failure

[peterbarker]

Bug report

Issue details

This branch abort()s the simulation if we skip an axis. If you are flying past the 60 second mark you can expect it to do so:

https://github.com/ArduPilot/ardupilot/compare/master...peterbarker:wip/ekf2-magfusion-issue?expand=1

https://github.com/ArduPilot/ardupilot/compare/master...peterbarker:wip/ekf2-magfusion-issue?expand=1 <- that branch shows the EKF2 issue I was muttering about yesterday.  My understanding is that our handling of each axis is supposed to be independent (assumed no coupling between them).  Those hacks seem to show that a failure on one axis will cause us to skip the next axis.  So if x fails we go straight to z.  You can test that in SITL with "arm throttle; rc 3 2000; rc 3 1500"... and then wait.  Suitable pauses required so you're actually flying, of course :-)
  You can expect it to abort at the 60 second mark when we actually force-fail the mag.

Paul: If we fail one, we should fail all three or only the affected axis.

Version
master

Platform
[ X ] All
[ ] AntennaTracker
[ ] Copter
[ ] Plane
[ ] Rover
[ ] Submarine

Airframe type
All

Hardware type
SITL

Logs
n/a

[peterbarker]

@priseborough fixed this :-)