regression in plane takeoff behaviour in 4.6

[tridge]

in 4.6.x we have introduced a bug in takeoff handling. We are not correctly handling an incorrect yaw estimate on takeoff. This is demonstrated in the following flight using CMAC_South, so the plane is pointing approximately south, with COMPASS_USE disabled:

the plane is flying to a loiter location behind the pilot, which is extremely bad
here is the same test with 4.5.7:

logs here:
http://uav.tridgell.net/TakeoffError/

[menschel]

As reported following #28352 there are three observed regressions in takeoff behaviour.

• The not excepted yaw reset by GPS leading to the course corrections mentioned above. EDIT: Not for Auto Takeoff Mission item, this turned out to be excessive crosswind deviation and takeoff does not track the path but only the heading.
• The continuous use of 100% throttle after TKOFF_THR_MAX_T is exceeded until TKOFF_ALT or the mission items takeoff altitude is reached. Previously it fell back to TECS controlled throttle based on the pitch angle.
• The level-off functionality can level out the plane below target altitude thus preventing takeoff to complete Issue Takeoff not completing under TKOFF_OPTIONS[0]=0 #28685 .

for reference ap460beta1_tkoff_issue.mp4 24-11-18_12-43-12.bin

[menschel]

The YAW issue is not present in auto takeoff mission item.
Measurements are referenced in the PR for the items I mentioned.
#28776