Unload meshes from the CPU automatically

[JMS55]

Objective

• Meshes are currently automatically uploaded to the GPU. However, once on the GPU, copies of all vertex data and indices are still kept around in RAM, which is a waste of memory.
• Furthermore, doing the Mesh->GpuMesh asset lookup every frame for rendering is a minor waste of CPU time.
• Supersedes Mesh: added static and dynamic mesh for less runtime memory usage #1782.

Solution

• Mesh no longer implements RenderAsset.
• A new system called transfer_meshes_to_gpu will take any entity with a Handle<Mesh> component, upload the mesh to the GPU and attach a GpuMesh, and then remove the handle.
  • If this handle is the last instance, refcounting will then delete the underlying mesh data.
• Render systems now query for GpuMesh directly.

TODO

• Fix examples
• Check that the GLTF loader isn't keeping a different copy of the data in memory
• Extend this PR to textures
  • Maybe keep RenderAsset around, but revamp it to convert handles instead
• Write changelog and migration guide

---

Changelog

This section is optional. If this was a trivial fix, or has no externally-visible impact, you can delete this section.

• What changed as a result of this PR?
• If applicable, organize changes under "Added", "Changed", or "Fixed" sub-headings
• Stick to one or two sentences. If more detail is needed for a particular change, consider adding it to the "Solution" section
  • If you can't summarize the work, your change may be unreasonably large / unrelated. Consider splitting your PR to make it easier to review and merge!

Migration Guide

This section is optional. If there are no breaking changes, you can delete this section.

• If this PR is a breaking change (relative to the last release of Bevy), describe how a user might need to migrate their code to support these changes
• Simply adding new functionality is not a breaking change.
• Fixing behavior that was definitely a bug, rather than a questionable design choice is not a breaking change.

[github-actions]

Example no_renderer failed to run, please try running it locally and check the result.

[vallrand]

Given that certain assets have to be kept in RAM anyway (shaders for recompilation when specializing, dynamic meshes that are mutated from the CPU side), asset itself might need to have an ability to be "unloaded" when consumed by internal system.
I would however try to look for ways to keep end user API consistent.

A new system called transfer_meshes_to_gpu will take any entity with a Handle component, upload the mesh to the GPU and attach a GpuMesh, and then remove the handle.
If this handle is the last instance, refcounting will then delete the underlying mesh data.

If Handle<Mesh> is how mesh is attached to entity, end user would expect to be able to call commands.entity(entity).remove::<Handle<Mesh>>(); later.

Also assets are usually managed separately (load/unload on new scene), so handle will never be removed (since it's kept inside some "resource"), And when removed, you might as well unload GPU data as well.
See bevy_asset_loader

[JMS55]

I'm going to close this for now and come back to this after the asset rework happens.