EKF2: fix yaw variance initialization

[bresch]

Solved Problem

By checking the yaw variance (error around the Down axis in local frame) before and after a yaw reset I realized that something was incorrect. For example I was getting this (from the pitch90 initialization unit test):
increase yaw var by 0.090000: 0.000335 -> 0.010335

Solution

Use SymForce to compute the equivalent quaternion variance from some yaw variance defined in NED
Now:
increase yaw var by 0.090000: 0.000340 -> 0.090340

Changelog Entry

For release notes:

Fix yaw uncertainty initialization
New parameter: -
Documentation: -

Test coverage

• unit test

[dagar]

Nitpick, maybe we should know from the signature we're only getting the lower triangle populated.

[bresch]

Good point

[bresch]

done

[bresch]

I'll also update the tilt covariance initialization auto-code soon (in another PR)