Threadgroup memory in metal mesh shaders

[inner-daemons]

Is your feature request related to a problem? Please describe.
Currently, due to very strange limitations with metal-rs, the mesh shader implementation in #8139 doesn't allow using threadgroup memory.

Describe the solution you'd like
Once metal-rs gets updated or objc2-metal is incorporated this should be addressed

Describe alternatives you've considered

Additional context
See also #8330, as passthrough shaders will still have issues after this is addressed