Function safe_arch::floor_m256

#[must_use]pub fn floor_m256(a: m256) -> m256

This is supported with target feature avx only.

Round f32 lanes towards negative infinity.

let a = m256::from([1.1, 2.5, 3.8, 5.0, -0.5, -1.1, -2.7, -3.0]);
let b = floor_m256(a).to_array();
assert_eq!(b, [1.0, 2.0, 3.0, 5.0, -1.0, -2.0, -3.0, -3.0]);