Re-enable Draco quantization

[ebogo1]

After #9841 gets merged, we will no longer quantize any attributes in the Draco worker. Some immediate tech debt from this PR:

• This is a temporary workaround while we switch to the draco3d npm module, but it should be doable to at least get some quantization benefits by packing compressed attributes into 16 bits. In the PR, c9ac980 has some progress towards getting the math to work in the shader but has inconsistent results with different models (i.e. with Specs/Data/Models/DracoCompression/BoxVertexColorsDracoRGBA.gltf Draco returns a null pointer even on the "NORMAL" attribute, which is unexpected).
• Some specs related to quantization were changed or deleted. These two commits should be reverted once quantization is added back: 1a3b697, 97c186c

Since this is a sizable regression, we should fix as much as we can in main ASAP and consider a follow-up 1.86.1 release.

related: #9787

[ebogo1]

Update: most of the specs are now back in main after #9904 got merged. The specs for PointCloud3DTileContentSpec.js are the only ones left, and should be added in once the Draco issue is fixed.

[ebogo1]

After #9908 is merged, all the specs will have been added back in. Once the Draco bug is fixed in the next release and we update the Draco version in CesiumJS, this issue can be closed.