per-joint segmentation works at joint face level

[DamienGilliard]

This PR concerns the modifications to the joint segmentator needed to have metrics for each joint face

It modifies the DFJointSegmentator:

• It outputs a list of lists of pointclouds, representing all the joints and the faces they contain. The structure is thus: [[face_1_joint_1, face_2_joint_1, ...], [face_1_joint_2, face_2_joint_2, ...], ..., [face_1_joint_n, face_2_joint_n...]]
• The list of faces has been registred to the reference, so we can accurately evaluate the face individually. This reflects the case where the faces might be correctly cut, but the whole joint badly placed in the beam/piece. Therefore we re-align the faces to the reference joint before evaluating them individually. Obviously all the faces of a joint follow the same transformation.
• The max correspondance distance parameter needed for the small ICP is exposed to the user, with a standard value of 5mm (assuming Rhino workspace is in meters)
  

when flattening the list of lists:

It also modifies a bit the DFCADSegmentator, because it also uses the same c++ back-end that has been slightly modified and thus requires some tweeking but only on the surface.

[DamienGilliard]

@eleniv3d , @9and3 , @mariklad This PR still needs the documentation (which I will do tomorrow, saturday, I promise), but the rest is working fine: we have the data to make a face-per-face analysis !
As always I shot myself in the foot for a few hours so it comes a bit late (it turns out deleting the .nuget folder at the same time as the .rhinocode one when you troubleshoot is a BAD idea, it cost me at least 2 hours )

I leave it as "Draft" until the documentation is done

[9and3]

@eleniv3d , @9and3 , @mariklad This PR still needs the documentation (which I will do tomorrow, saturday, I promise), but the rest is working fine: we have the data to make a face-per-face analysis ! As always I shot myself in the foot for a few hours so it comes a bit late (it turns out deleting the .nuget folder at the same time as the .rhinocode one when you troubleshoot is a BAD idea, it cost me at least 2 hours )

I leave it as "Draft" until the documentation is done

Hello @DamienGilliard thank you a lot for this push! I'll take care of the documentation and few fixes no worries, many thanks again it looks already very nice! 👐

[9and3]

Made some tiny cleaning, I also believe that there are parts in the segmentation backend that could be refactored to improove the performance but this is for another day :)

the major point I modified is this one:

when flattening the list of lists:

I fixed this @DamienGilliard (and for later @eleniv3d ). Now faces export ghtrees by converting it to ghtree, it's acutally quite simple:

import ghpythonlib.treehelpers as th
gh_tree = th.list_to_tree(py_nlist)

Maybe @eleniv3d in the computational error component you will have to reconvert it and you might have some surprise to deal with the input types (since this can be now a ghtree and hot only ghlist). In any case I am merging this, thanks again @DamienGilliard !