math: speed up bessel functions on AMD64

math: speed up bessel functions on AMD64

J0-4            71.9ns ± 1%  54.6ns ± 0%  -24.08%  (p=0.000 n=20+18)
J1-4            71.6ns ± 0%  55.4ns ± 0%  -22.60%  (p=0.000 n=19+20)
Jn-4             153ns ± 0%   118ns ± 1%  -22.71%  (p=0.000 n=20+20)
Y0-4            70.8ns ± 0%  53.9ns ± 0%  -23.87%  (p=0.000 n=19+19)
Y1-4            70.8ns ± 0%  54.1ns ± 0%  -23.54%  (p=0.000 n=20+20)
Yn-4             149ns ± 0%   116ns ± 0%  -22.15%  (p=0.000 n=19+20)

Fixes #16889

Change-Id: Ie88496407b42f6acb918ffae1226b1b4c0500cb9
Reviewed-on: https://go-review.googlesource.com/28086
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>