gomath: fix negative pitch singularities in RPY and ZYX conversion

[Rimas2200]

Fixes #389.

Both matrix-to-Euler conversions checked for +pi/2 twice, leaving the negative-pitch branch unreachable. That branch also used m->y.z to recover roll, although this element is zero at pitch = -pi/2.

Changes

• Fix the negative-pitch check in go_mat_rpy_convert() and go_mat_zyx_convert().
• Recover roll using -atan2(m->y.x, m->y.y), keeping the existing convention of yaw = 0 at the singularity.
• Add tests for both poles, rotations inside and outside the singularity tolerance, ordinary rotations, and quaternion conversions with q and -q.
• Check orientation using independent matrix multiplication.

Testing

• uspace build and pycheck passed (make -O -j4 default pycheck V=1).
• tests/posemath/gomath-euler-singularities passed.
• The targeted conversions also passed with float and long double.

[grandixximo]

Pointing at 2.9, again we might want it on master first @andypugh @BsAtHome ?