Function safe_arch::sub_horizontal_m256

#[must_use]pub fn sub_horizontal_m256(a: m256, b: m256) -> m256

This is supported with target feature avx only.

Subtract adjacent f32 lanes.

let a = m256::from([8.0, 17.0, 6.0, 5.0, 4.0, 23.0, 2.0, 1.0]);
let b = m256::from([0.0, 2.0, 4.0, 8.0, 16.0, 32.0, 64.0, 128.0]);
let c = sub_horizontal_m256(a, b).to_array();
assert_eq!(c, [-9.0, 1.0, -2.0, -4.0, -19.0, 1.0, -16.0, -64.0]);