[PATCH 4/5] add volk_32f_s32f_x2_convert_8u kernel

[PATCH 4/5] add volk_32f_s32f_x2_convert_8u kernel

This adds a kernel that performs conversion from float to uint8_t
using a scale and a bias, according to out = in * scale + bias. The
output is clamped to the interval [0, 255] and rounded to the nearest
integer.

The kernels are implemented mirroring those for
volk_32f_s32f_convert_8i, but there is an additional avx2_fma kernel,
since it makes sense to perform the scale and bias conversion using FMA
if it is available.

Signed-off-by: Daniel Estévez <daniel@destevez.net>
Gbp-Pq: Name 0004-add-volk_32f_s32f_x2_convert_8u-kernel.patch