EKF possible height reset bug

[dagar]

Multi-EKF in SITL (3 identical IMUs, 2 identical mags)

All estimators experienced a baro hgt timeout, but due to timing some reset to GPS, some to baro. Of the 4 instances logged (logger limit) one experiences a 200 meter jump in Z during the reset. Unfortunately due to the low logging rate at the time we can't directly see which instance reset to baro vs GPS.

https://logs.px4.io/plot_app?log=b3c6ce24-aad7-4040-8c7e-cbcbf98b1fb1

[priseborough]

The logs show that that index 3 instance initialises to the incorrect height origin.

Cause TBD.