Implement PSLG
Issue #6
new
mshr currently uses CGAL::Snap_rounding_2 to compute the PSLG and avoid roundoff errors.
An unfortunate effect is that all vertices are moved (by at most pixel_width/2). It is also probably an overkill since roundoff errors is not really problem in the PSLG (short edges on the other hand, still need to be handled).
The snap_rounding package has shown to be the bottle neck when the number of vertices in the PSLG increases.