Something is amiss in the reference frames whose axis of rotation varies

[eggrobin]

It was noted in #3609 that we had an abnormal discrepancy between the motion computed in the inertial frame and that computed using the geometric acceleration in the rotating-pulsating frame, but that this discrepancy was there in the MEO frame already.

This discrepancy is absent in the frame of the lunar surface. Interestingly @pleroy found that it is also absent when the Sun-Earth-Moon system is made planar; thus the issue is not just nonuniform rotation (the orbit of the moon is still eccentric in that case), but variation of the axis of rotation.

[pleroy]

The current theory is that this computation is incorrect. What we should be computing is the spin angular velocity of a rigid body (the frame) but we are using the formula for orbital angular velocity; Wikipedia even has a note explaining that this is wrong.

[eggrobin]

Fixed by #3625.