helical_nstart option for refinement is broken

[robertbuecker]

The --helical_nstart option of relion_refine does not work in the current main version - instead of reverting the twist back to the n-start convention, it adds a multiple of 360 on top (e.g. a 2-start helix with 180.5° 1-start twist becomes 721° instead of 1° in 2-start convention). For my application - a C2_1 amyloid, where rot priors don't work out of the box - using --helical_nstart=2 hence made matters even worse. The problem results from a typo in helix.cpp (line 4425). After fixing it, it works well (at least for the C2_1 amyloid in both right-handed and left-handed config) and improves the reconstruction significantly.