multi_radixsort.comp fails to compile on AMD graphics

[pavelsevecek]

The compilation of the compute shader fails on AMD Radeon 780M Graphics. The error is on the line 74: https://github.com/MircoWerner/VkRadixSort/blob/96178e0515292f96de112fea0c1b469416388c7e/multiradixsort/resources/shaders/multi_radixsort.comp#L74C16-L74C75

subgroupBroadcast is using gl_SubgroupID which isn't a compile-time constant.

The compiler further complains about using atomicAdd on the bin_flags variable, but that can be easily fixed by flattening the array.