fix leak in extract_polygonal_prism_data

[v4hn]

The inlier-check worked before Alexandru broke it in 2012.
(see 6149729) This last test in the
raytracing code is actually already the first one that is tested for.

[v4hn]

The current version of pcl produces this weird bug when asked to extract all points within the (blue) convex hull
The pointcloud here is a 2d-projection of a scene from which the convex table hull is extracted onto the table plane.
The two purple points mark the beginning and end point of the convex hull. The red points are all points which should be on or above the convex hull.
With this patch the correct set of inliers is computed:

[taketwo]

What about these: #869 #552?

[v4hn]

I agree with @gedikli in #869. There never was a problem with explicitly closed polygons,
because in the relevant loop iteration the condition 0<0 will always turn out to be false.
That's also the reason why people where able to work around this bug by using explicitly closed polygons: in this case the superfluous test in the end was not harmful anymore.
With this bugfix, implicitly as well as explicitly closed polygons are supported again and both issues you referenced can be closed.

[taketwo]

@v4hn Great! Thanks for bringing this up again.