Added support for glTF 1.0.1 normalized to quantizeAttributes

[lasalvavida]

Adds an optional flag to use the glTF 1.0.1 normalized property for higher precision decode matrices.

[lasalvavida]

Used to generate the test model for CesiumGS/cesium#4247

@lilleyse, can you look at this when you get a chance?

[lilleyse]

@lilleyse, can you look at this when you get a chance?

Yes I'll check this out soon.

[pjcozzi]

CC @lexaknyazev - glTF 1.0.1

[pjcozzi]

I'm not following why this is needed.

[lasalvavida]

This is one of the advantages of having glTF 1.0.1. By adding the normalized flag to a quantized accessor, the decode matrix doesn't need to be stored as 1.0/65535.0, since this is done automatically by normalizing the data. This gives you the option to generate quantized models that decode with higher precision by not rounding in the glTF.

The rounding is done by the GPU when it normalizes, so you automatically get the precision of the system you're running on.

It also shrinks the glTF asset since the decode matrix won't be full of enormous floating point strings anymore.

[lilleyse]

+1 from me

[pjcozzi]

OK, does this still strictly conform to the extension or does that need an update?

[lasalvavida]

No updates needed, this strictly conforms to the extension.