Multiple objects can have the same id. Using that id to compare the render
objects makes the sort function unstable. Use the webgl object's index in the
render list instead. That number is guaranteed to be unique.

This fixes #2966.

Prevent BufferGeometry custom attributes's array from being deleted by setting geometry.attributes.name.dynamic = true

Its tempting to go on, but would stop here to make sure there's no regression.
Also, will have to port these changes over to WebGLRenderer2

This seems to get 15fps faster than the non BufferGeometry counterpart.
Note that buffergeometry.dynamic = true is required for updates
Also cleaned up WebGLRenderer

Use

geometry.attributes.index.needsUpdate
geometry.attributes.position.needsUpdate
geometry.attributes.normal.needsUpdate
geometry.attributes.uv.needsUpdate
geometry.attributes.color.needsUpdate
geometry.attributes.custom.needsUpdate

instead of for usual geometries

geometry.verticesNeedUpdate
geometry.elementsNeedUpdate
geometry.uvsNeedUpdate
geometry.normalsNeedUpdate
geometry.colorsNeedUpdate
geometry.tangentsNeedUpdate

- fixes #3017 and possibly #3084

Seems like @gero3 is MIA. We'll try again on the next cycle.

Ok, lets give this a push :) @gero3 webgl_lensflares.html seems to be broken.

make Matrix4.getPosition() deprecated and replace all usages in core library with Vector3.getPositionFromMatrix()

Renamed Vector3 projectPoint() to applyProjection() as per @WestLangley suggestion in bb873d16.

@WestLangley I've changed most of the vector.applyMatrix4() to vector.projectPoint().
Feel free to change back the ones that aren't needed (or let me know which ones can need to be changed back and I will do it).

Extra feature detection tests, to know what to load when using library

e.g. 
If no compressed texture support, load regular textures etc
If no float texture support, do multi pass rendering and pack float into rgba
If no dervitives do something else...

Removed Vector* add() to addVectors() and addSelf() to add(). Same with sub, cross, min, max, lerp... Renamed Matrix* and Quaternion multiply() to multiplyMatrices(), multiplyQuaternions() and multiplySelf() to multiply(). See #2860.
Good thing we have a ton of examples to make sure nothing is broken. The unit tests also helped. However, I suspect I've left something unchanged.

Frustum.contains -> intersectsObject, containsSphere->intersectsSphere, removed Frustum.containsAnyPoints, .transform.  Update unit tests accordingly.

Renamed .multiplyMatrix* to .applyMatrix* and .multiplyQuaternion to .applyQuaternion as suggested by @WestLangley in #2860.

Makes the code so much nicer imho...

Fixed inconsistent precision for lensflares and sprites shaders causing troubles on mobile (thanks to @phoboslab).

See #2769

Renamed .deallocate() to .dispose() as discussed in #2760 and fdb737d0.

Still unsure of whether to rename it to dispose instead.
I guess a problem of the world deallocate is that you're unsure wether it is dealocate or deallocate.

See #2624

This will need some more thought, before we didn't really do anything about BufferGeometry deallocation.

It's hard to test because, I think, chrome is now automatically deallocating stuff on it's own
@alteredq I've done a combined deallocateGeometry and some things here and there would need to be refactored but... what do you think?