Blender Git Loki

Add inlined custom lnors consistency check/fixup code in BMesh version of `loops_calc_normals`

Note that stashing all clnor computations in a single func might not be optimal,
but it allows to reduce compute overhead as much as possible (it's a desing decision
in the end).

Also, looks like func used to create smooth fans from a set of custom lnors is not working
100% correct - new checking code did found some diff in lnors of same smooth fans... :/

Add DM callback to get lnors spaces (needed by modifiers generating custom normals!)

When converting custom normals to CD data, interpret NULL vectors as NOP ones.

i.e. if you give a (0, 0, 0) vector, you will always get auto-computed normal...

Fix various little issues and some corner-cases bugs...

Rework a bit `BKE_mesh_normals_loop_custom_set()`, should be more robust now.

Some cleanup, and add same 'proof-of-concept' code to bmesh,
so that we can visualize custom loops normals in 3DView!

Still many ugly things here, esp. in includes...

Enhance lnor space handling (by taking into account the 'cone case').

In other words, we now have two plain angles, and custom normals are encoded as
factors for those two angles.

Add a 'clear custom split normals' op (similar to those for skin or mask CD layers).

Fix stupid handling of clnors_data in editderivedmesh.

There was a mismatch between index/offset, and we need to handle both clnors from dm and from actual bm data...

Add inlined-checks/fix of clnors data consistency to BKE_mesh_normals_loop_split() as well...