Export glTF skins as a Gltf struct

[barsoosayque]

Objective

• Make skin data of glTF meshes available for users, so it would be possible to create skinned meshes without spawning a scene.
• I believe it contributes to bevy_gltf improvements to decouple it from scene creation #13681 ?

Solution

• Add a new GltfSkin, representing skin data from a glTF file, new member skin to GltfNode and both skins + named_skins to Gltf (a la meshes/nodes).
• Rewrite glTF nodes resolution as an iterator which sorts nodes by their dependencies (nodes without dependencies first). So when we create GltfNodes with their associated GltfSkin while iterating, their dependencies already have been loaded.
• Make a distinction between GltfSkin and SkinnedMeshInverseBindposes in assets: prior to this PR, GltfAssetLabel::Skin(n) was responsible not for a skin, but for one of skin's components. Now GltfAssetLabel::InverseBindMatrices(n) will map to SkinnedMeshInverseBindposes, and GltfAssetLabel::Skin(n) will map to GltfSkin.

Testing

• New test skin_node does just that; it tests whether or not GltfSkin was loaded properly.

Migration Guide

• Change GltfAssetLabel::Skin(..) to GltfAssetLabel::InverseBindMatrices(..).

[github-actions]

Welcome, new contributor!

Please make sure you've read our contributing guide and we look forward to reviewing your pull request shortly ✨

[yrns]

I don't know that this is a problem per se, but I noticed that the error message "Unexpected child in GLTF Mesh" is from a older commit (removed in #13707). Which raises concern that this set of changes doesn't reflect more current work or there was a merge problem. (I'm not sure!) I only noticed because I was confused about "mesh" given the context. Maybe double check?

[barsoosayque]

Which raises concern that this set of changes doesn't reflect more current work or there was a merge problem.

It's true that some work in this part of the codebase may be omitted because I ported these changes from 0.12 to current main and. But I double checked and everything seems to be fine, the error about gltf mesh child is basically irrelevant anyways according to the pr that removed it ?

[alice-i-cecile]

This is a reasonable step forward :) Thanks for putting in the work to get this together.