KHR_draco_mesh_compression: Clarify accessor details and indices resolving.

[McNopper]

Please clarify, which values the accessor should have:

count: Is this the value for resolved or unresolved values? From my understanding it should be the resolved value.

componentType: If the uncompressed data is decoded, the result has the given component type and no conversion is needed? I assume no conversion is needed.

type: If the uncompressed data is decoded, the number of elements are "correct" and no conversion is needed? Like above, I assume no conversion is needed.

Finally, it should be explained, how indices have to be resolved.

[pjcozzi]

@FrankGalligan?

[jackcaron]

@McNopper I got it working using the same values as without Draco compression. And don't forget the min/max for the position accessor.

[McNopper]

We also have it working in our engine, but it should be better explained and specified.

[prideout]

This issue was partially resolved by #1343 but I think the spec language could still be improved. For example, the following sentence doesn't specify which fields of the accessors should be replaced:

A loader must use the decompressed data to fill the accessors

Also, I would suggest changing:

must match the decompressed data.

to:

must match the decompressed data, which is potentially different from the source data.

The distinction between decompressed data and source data isn't obvious to folks who are reading the spec for the first time.

Also, as a public service announcement, please note that many conformance models currently do not comply with the latest Draco spec language, see https://github.com/KhronosGroup/glTF-Sample-Models/issues/234 for details.