Feature ID changes from EXT_mesh_features spec updates

[ptrgags]

This PR targets a staging branch, feature-id-changes

the EXT_mesh_features extension is under some revisions, being split into EXT_mesh_features and EXT_structural_metadata.

This PR doesn't cover everything, but it's a start. It includes:

• the FEATURE_ID_n -> _FEATURE_ID_n attribute semantic
• added nullFeatureId and featureCount to ModelComponents.FeatureIdXxx (though these are not yet used for anything)
• feature ID textures now have the texture info moved into a texture property.
• Parse feature IDs from a multi-channel texture

Sandcastle for multi-channel Feature IDs -- some of the feature ID textures are a bit contrived, I need to find a better way to check them for correctness. I also need to try picking/styling with the one texture that has null feature IDs.

[cesium-concierge]

Thanks for the pull request @ptrgags!

• ✔️ Signed CLA found.

Reviewers, don't forget to make sure that:

• Cesium Viewer works.
• Works in 2D/CV.

[ptrgags]

@lilleyse the code is now ready for review. Note that I want to make two Sandcastles for testing:

• Sandcastle link Update the sandcastle from the description to have better visualizations for the larger feature ID values -- the comments explain what should happen. Everything works fine except the RGBA one which fails due to precision issues (as expected).
• Make a sandcastle for picking the 7th feature ID texture to check if the null feature ID code works correctly -- not actually possible since there's no property table, see Model EXT_mesh_features: handle feature IDs without a property table #9884

[ptrgags]

in that first sandcastle, it seems that fractal pattern I found yesterday happens only in the case where I use all 4 channels (RGBA for the first, AGBB for the second). Somehow the gradients in each channel + precision limitations combine to make a fractal pattern.

These precision issues were to be expected given that highp int is actually represented as a float32 in GLSL 1.00. All the other cases with 1-3 channels work fine.

[lilleyse]

Looks good! Just a couple comments. I tested picking in the 3D Tiles Next sandcastles as well as pnts, i3dm, and b3dm with enableModelExperimental: true.

[ptrgags]

@lilleyse just a note, I do plan to clean up this changelog before we merge the staging branch. it'll probably be best to make a list of sub-bullets (like we did when we first introduced 3D Tiles Next) but that'll be easier once everything is in place.

[ptrgags]

@lilleyse updated!