Remove euler angles from attitude setpoint

[Jaeyoung-Lim]

Solved Problem

The VehicleAttitudeSetpoint message has two different attitude setpoint fields.
a) The desired attitude quaternion q_d

PX4-Autopilot/msg/VehicleAttitudeSetpoint.msg

Lines 3 to 10 in 75ce550

	float32 roll_body # body angle in NED frame (can be NaN for FW)
	float32 pitch_body # body angle in NED frame (can be NaN for FW)
	float32 yaw_body # body angle in NED frame (can be NaN for FW)
	float32 yaw_sp_move_rate # rad/s (commanded by user)
	# For quaternion-based attitude control
	float32[4] q_d # Desired quaternion for quaternion control

b) Euler angles roll_body, pitch_body, yaw_body

I think the initial intention was that to use the euler angles for logging, and the quaternions for the controllers. However, this seems to have gotten quite abused, and the fixed wing controllers especially use it as a primary place for storing setpoints.

This has the following problems

• When both euler angles and quaternions are in the message, but the two contain different attitude information, there is no information in the message telling which one is the valid message to be used.
• The way currently the euler angles are used make the code become quite hard to read. Especially the fixedwing position controller, since the euler angles of global variable _att_sp is modified everywhere in the code. This makes it hard to grasp where the setpoints are changing and which euler angles are actually sent to the attitude controller
• The field names (e.g. roll_body) has nothing to do with the body frame at all.
• When using the attitude setpoints from an external process (e.g. ROS2), it is not clear which one will be used, or how limited information can be provided. For example, the sender might invalidate all setpoints and set a roll_body. This is not a valid setpoint and therefore should simply not be allowed from the message definitions itself.

Fixes #23481

Solution

• Remove euler angles from attitude setpoint, and compute the euler angles locally when needed

Changelog Entry

For release notes:

Feature/Bugfix Removed euler angles from attitude setpoint message.

Test coverage

• Test plane in SITL

[dagar]

Let's do it, this is long overdue.

[Jaeyoung-Lim]

Not sure if the CI failures are relevent. SITL tests say that the tests passed, but still fails anyway.

[bresch]

Nice

[sfuhrer]

Not sure if the CI failures are relevent. SITL tests say that the tests passed, but still fails anyway.

No they're unrelated, see #23457

[Jaeyoung-Lim]

@bresch Sorry, I had a format error during the rebase

[sfuhrer]

This was really long overdue, thanks a lot!
I just love my consts too much - would you mind going through and const everything you can? I stopped making comments after a while, there's lots of them.

[Jaeyoung-Lim]

@sfuhrer Addressed your comments

[sfuhrer]

Looks all clean now, thanks!