Support kernel dispatch by total number of threads instead of threads per threadgroup

I think the real fix may be to switch from [`dispatchThreadgroups`](https://developer.apple.com/documentation/metal/mtlcomputecommandencoder/1443138-dispatchthreadgroups?language=objc) to [`dispatchThreads`](https://developer.apple.com/documentation/metal/mtlcomputecommandencoder/2866532-dispatchthreads?language=objc) by default since all the devices we support support nonuniform threadgroup sizes.

However, this would be a very big (potentially breaking) change.

_Originally posted by @christiangnrd in https://github.com/JuliaGPU/Metal.jl/issues/496#issuecomment-2543146126_