Perform autocal with EKF mag bias estimate

[bresch]

requires: #23584

Solved Problem

A separate magnetometer bias estimator was added as it was not safe to activate the mag bias estimation of earlier EKF2 versions. This is now solved and having 2 different estimator affecting the mag biases adds a lot of unnecessary complexity.

Solution

Use EKF2 mag bias estimates only (also for autocal).

Changelog Entry

For release notes:

New parameter: none (removed MBE_ENABLE and MBE_LEARN_GAIN)
Documentation: TBD

Test coverage

Tested on a Pixhawk 4 (internal mag only).
Left: autocal using EKF2 mag bias estimate ----------------- Right: 6-sides calibration

[dagar]

This might be controversial, but I think we should keep the possibility to establish an initial mag cal (entirely preflight) if mag is completely uncalibrated. It could be sufficiently strict so that regular users don't accidentally slip through and avoid the full mag cal dance, but I think it's a nice convenience for bench testing. I think we can be sufficiently strict so that you actually have to wave things around and get a decent bias estimate or perhaps hide this behind a parameter.