Start experimenting with 4-byte assignment using workgroup size as 1 byte via local_i...

Start experimenting with 4-byte assignment using workgroup size as 1 byte via local_invocation_id and then 3 dimensions in dispatch as 3 more bytes via workgroup_id.x/y/z, all concatenated into global_invocation_id. Looping in compute shaders is also OK versus fragment shaders, so iterate within the shader instead of reading out every frame with CPU.