jolt physics: to_jolt to use get_rotation_quaternion

[vmbaillie]

The orthanormalized basis transform from set_transform sometimes has a negative determinant, so instead of calling orthonormalize within that routine, it's called within get_rotation_quaternion.

Fixes #103408

[vmbaillie]

There may be other places to reduce calculation repetition, best to get an expert to review before any merge

[fire]

We tend to use get_rotation_quaternion for basis to quaternion conversion. So this seems correct.

Part of the animation team.

[mihe]

I'm not entirely opposed to merging the changes to the conversion function, since it is technically safer this way, but #103440 still seems like a more holistic approach to me, so let's resolve the discussion there first.

However, as hinted at by @vmbaillie already, there are more places than JoltBody3D::set_transform where redundant orthonormalization should be removed in this case, so this is not ready to be merged either way.

[mihe]

With #103440 having been merged, I would consider this PR to be superseded, as the additional work being done in Basis::get_rotation_quaternion compared to Basis::get_quaternion, meaning orthonormalization and ensuring correct handedness, is entirely redundant.