More performance updates for WebGPU mode

[davepagurek]

Resolves #8471

Changes

• All drawBuffers calls are batched into a single render pass now, since it should be safe to do so as long as we still end the last render pass when we switch from drawing to framebuffers
• We no longer draw all uniforms into a single buffer every frame:
  • I've split up uniforms into multiple chunks, so we can change behaviour on a per-chunk basis rather than all or nothing
  • For most chunks, we now cache the last bound buffer so that we don't have to rewrite it when the value doesn't change
  • Writing is still a bottleneck for chunks that update for basically every call. For those, we now write to a part of a big memory-mapped buffer. Right now I'm doing this for position-related data and also color (but not other material settings) since those feel most likely to be updated per object.

PR Checklist

• npm run lint passes
• Inline reference is included / updated
• Unit tests are included / updated