AVX-512: Optimized vector shuffle for v16f32 and v16i32 types.