Improvements in point association

[DamienGilliard]

This PR concerns improvements in the point association and an adaptation in the angle threshold for considering a segment "associable".

It is minor in code, but quite visible in results

Change n° 1

DFMesh::IsPointOnFace():
* The coordinates u, v on the triangle face are not really barycentric coordinates. The comment in the code has been adapted.
* Given the nature of u, v, a small deviation can lead to large difference in 3D coordinates. For that reason, the association threshold is divided by 100 for more precise manipulation

Change n° 2

DFSegmentation::AssociateClustersToMesh() and CleanUnassociatedClusters():
* The angle between the face normal and the line that joins the face center and the cluster center is taken into account. Ideally this angle must be very close or equal to 90°. A condition states that the cosine of this angle must be smaller than the angle threshold.

[9and3]

Hello @DamienGilliard ! Thanks for this PR, I see it is still a draft, can we merge this as a temporary patch already? Thanks!

[DamienGilliard]

I got a bit lazy. I'll document it before 12:00 today and then you merge ?
Sorry I should have done that last week.

[9and3]

fine for me to merge this little patch. I still belive the python segmentators need some refactoring but we'll do this in another PR! Thanks for this one!