RotationNR fails to maintain consistency when converted to rotation angles and back

[madhephaestus]

RotationNR sometimes produces invalid rotations after a conversion from rotation angles and back again. The failure could be in the quaternion conversion, the rotation angle conversion or maybe a separate bug in each. Additional unit test cases could and should be added so all conversion space and types of absolute rotations should be checked.

[madhephaestus]

It would seem the new library is not producing rotation angles from quaturnions in a way that is API compatible with the previous system.