Fix quaternion gimbal lock

[vr4guid]

fix _dcm_to_euler calculation when pitch is -pi/2

[meee1]

all my sources indicate it was correct before this change.

[vr4guid]

This corrected the issue on our side.

[robertlong13]

This corrected the issue on our side.

How? Isn't this a non-functional change? phi is set to 0 in the previous line. Now, that raises the question of what the - phi and + phi were ever doing in these two places to begin with, but this change can't possibly have done anything, unless I'm gravely mistaken (and I might well be; I keep reading it over and over).