Maximal rank which is understood by the CUDA kernel transpose.
Referenced by syten::CudaDenseTensorImpl::cuda_transpose_kernel().