Updated the way WebGPU dynamic UBs are used for rendering

[mvaligursky]

Fixes #6315 based on this comment: #6315 (comment)

• mesh bind group is split into two, one for a uniform buffer, and one for the rest of resources (textures and storage buffers and similar)
• this allows us to re-use the same bind ground from the dynamic uniform buffer for all meshes using the same dynamic buffer (per UB size). When a UB buffer jumps to a different per frame bump allocated buffer, it would simply use it's existing bind group instead of creating a new one
• a good example to demonstrate this is the `hierarchy' engine example, where due to large amount of meshes rendered in different order over time, many of them would need to allocate a new bind group. This does not happen now at all.
• note that this does not solve the case where a new bind group needs to be created due to assigning a different texture to it, as that is unavoidable

Separately, the order of bind groups in the layout is sorted the opposite way:

• original order: mesh -> view
• new order: view -> mesh -> mesh_ub
  This is a recommended way to sort, from the least frequently updated to more frequently updated, to allow the underlying platform to skip all follow up bind groups to be re-set on their device.

Few engine examples that use WGSL directly had to be change to work with this, as there is no automatic bind group allocation yet. This is a non-breaking change as those are not public APIs yet.