Add UV channel selection to StandardMaterial

[geckoxx]

Objective

• The StandardMaterial always uses ATTRIBUTE_UV_0 for each texture except lightmap. This is not flexible enough for a lot of gltf Files.
• Fixes Gltf texture not rendering properly #12496
• Fixes Loading a glTF scene doesn't use the correct UV map for ambient occlusion #13086
• Fixes Incorrect UV map used when using Lightmap component #13122
• Closes Gltf loader reads texCoord field of texture #13153

Solution

• The StandardMaterial gets extended for each texture by an UvChannel enum. It defaults to Uv0 but can also be set to Uv1.
• The gltf loader now handles the texcoord information. If the texcoord is not supported it creates a warning.
• It uses StandardMaterial shader defs to define which attribute to use.

Testing

This fixes #12496 for example:

For testing of all kind of textures I used the TextureTransformMultiTest from https://github.com/KhronosGroup/glTF-Sample-Assets/tree/main/Models/TextureTransformMultiTest
Its purpose is to test multiple texture transfroms but it is also a good test for different texcoords.
It also shows the issue with emission #13133.

Before:

After:

[geckoxx]

Added the same support for clearcoat textures and using shader defs instead of material flags.

[bugsweeper]

@geckoxx You should add "Closes #13153" to your description

[robtfm]

nice. clean code and provides a good base to extend to more UVs in future.

we could potentially select which texcoords to put into our uv attributes based on usage rather than just taking 0/1, but if we do extend to more uvs that would be unnecessary anyway, and it's a clear improvement as is.

[robtfm]

just a quick note about the shader defs - we could alternatively use indexes/bits in the material data to specify the uv indexes and avoid a proliferation of shader variants.

i think defs are probably better because

• it will have no cost in the base case where only 1 set of texcoords are used
• if you do need to use multiple sets, you likely have them all in the same locations, so only one variant would be needed
• if they are not in the same locations, if you control the assets and you care enough, you can modify your assets to align them

whereas using data in the material would increase its size having a (marginal) effect on all uses of StandardMaterial, whether you want the flexibility or not.

[alice-i-cecile]

Thank you to everyone involved with the authoring or reviewing of this PR! This work is relatively important and needs release notes! Head over to bevyengine/bevy-website#1313 if you'd like to help out.