See discussion in #2613

Finally I just left it behind a flag. I was not able to get as good results for TextGeometry as with unweighted method and split materials for side and front faces.

I wonder how I got that nice look as in screenshot I posted in the issue. It still does look better with weighted than unweighted but nowhere so close as in that screenshot. Either I did some mistake when testing before or there is some bug in implementation now.

Not yet 100% sure about all the details but this is something that should be useful for making helpers prettier.

Not tested yet, but it should work in the same way like vertex colors in Ribbons: add one Vector3 with normal per vertex to "geometry.normals" array.

It's then up to the application layer to fill in proper values. It's use case specific how normals for triangle strips would get computed - for simple ribbon meshes it could be very similar like smooth normal computation for regular Mesh objects.

If degenerate triangles are used for pseudo-connected single ribbon mesh planes, application would need to take into account vertex adjacency.

Experimenting with "geometry.buffersNeedUpdate".

This will avoid ending up with arrays full of nulls.

- Add parameter for vertex skimming
- Minimize square root calculations

- When a Face4 have duplicated vertices,
Geometry.mergeVertices() converts it into a Face3
- Face3 now copies normal, color, materialIndex, vertexColors
from the original Face4

Added dynamic updating of BufferGeometry. Added BufferGeometry.applyMatrix. Added Matrix4.multiplyVector3Array.

CTMLoader now keeps references to typed arrays used to initialize GL buffers. These are later on deleted in WebGLRenderer (by default, if BufferGeometry.dynamic flag is not set).

Fixes #2166

(this commit should go to `unflipped` branch, it assumes straightforward UVs)

allows a reduction to a face3 in Geometry's .mergeVertices(). so finally the
LatheGeometry example in subdivision surfaces is finally working as
intended!

(Phewww!) I'm sure there are still things I missed...

Should fix issue #1647

This adds Vector3.getRotationFromQuaternion and changes Object3D.applyMatrix to use Matrix4.decompose.

Also added SceneUtils.attach / detach and cleaned up unused vector parameter in Geometry.applyMatrix.

Default computation of morph normals for all models with morphs was taking unnecessary time and memory.

Also made it less resource intensive to recompute morph normals.

It seems to work, though not sure yet how to handle some things. Also for the moment it's implemented just for Lambert material.

As per discussion in #1170

Work in progress. For the moment works just with CTMLoader. Still not sure where some pieces of code would go.

Upsides:
	- initialization time cut down significantly (1.8 seconds => 0.7 seconds for CTM example)
	- much smaller GL buffers thanks to fully used indexing (Walt has 6x less vertices)
	- consequently much smaller memory footprint all around, both for main and GPU memories,
	  which means less swapping and better vertex caching

Downsides:
	- can't have anymore sharp per-face effects like flat shading (at least if mesh is not constructed in a way where each face has own vertices)
	- can't have multiple materials per mesh
	- no more easy manipulation of anything (for the moment actually no runtime manipulation at all, typed arrays are just thrown away, dynamic geometries code path not yet implemented)

Bad things (and hard to debug) can happen that affect Ray.

Do not use for anything, this is just snapshot of a test in progress. I solved quite a few issues, but many things are still clumsy and broken.