Function safe_arch::sub_horizontal_m128

#[must_use]pub fn sub_horizontal_m128(a: m128, b: m128) -> m128

This is supported with target feature sse3 only.

Subtract each lane horizontally, pack the outputs as a then b.

let a = m128::from_array([10.0, 20.0, 30.0, 45.0]);
let b = m128::from_array([100.0, 200.0, 300.0, 450.0]);
let c = sub_horizontal_m128(a, b).to_array();
assert_eq!(c, [-10.0, -15.0, -100.0, -150.0]);