* vulkan: combine duplicated fastdiv functions, rename the one optimizing small divs * remove one more fastdiv
* vulkan: Workaround compiler bug in conv2d coopmat2 path * apply same workaround to CONV_3D * Apply suggestion from @jeffbolznv
* vulkan: support CONV_3D This is a pretty direct port of conv2d_mm.comp to CONV_3D, done by codex and cleaned up by me. * disable slower perf tests