image: optimize bounds checking for At and Set methods

image: optimize bounds checking for At and Set methods

Use a subslice of the pixel data to give the compiler hints
for bounds checking. Only do this for image formats that
require 4 or more slice reads/writes.

See #27857 for discussion of small cap sizes.

name                   old time/op    new time/op    delta
At/rgba-8              18.8ns ± 2%    18.5ns ± 1%   -1.49%  (p=0.026 n=10+10)
At/rgba64-8            22.2ns ± 2%    21.1ns ± 3%   -4.51%  (p=0.000 n=10+10)
At/nrgba-8             18.8ns ± 2%    18.7ns ± 2%     ~     (p=0.467 n=10+10)
At/nrgba64-8           21.9ns ± 2%    21.0ns ± 2%   -4.15%  (p=0.000 n=10+9)
At/alpha-8             14.3ns ± 1%    14.3ns ± 2%     ~     (p=0.543 n=10+10)
At/alpha16-8           6.43ns ± 1%    6.47ns ± 1%     ~     (p=0.053 n=10+10)
At/gray-8              14.4ns ± 2%    14.6ns ± 5%     ~     (p=0.194 n=10+10)
At/gray16-8            6.52ns ± 1%    6.55ns ± 2%     ~     (p=0.610 n=10+10)
At/paletted-8          4.17ns ± 1%    4.21ns ± 2%     ~     (p=0.095 n=9+10)
Set/rgba-8             39.2ns ± 2%    40.1ns ± 4%   +2.45%  (p=0.007 n=10+10)
Set/rgba64-8           46.2ns ± 3%    43.3ns ± 3%   -6.11%  (p=0.000 n=10+10)
Set/nrgba-8            39.2ns ± 1%    39.7ns ± 5%     ~     (p=0.407 n=10+10)
Set/nrgba64-8          45.6ns ± 3%    42.9ns ± 3%   -5.83%  (p=0.000 n=9+10)
Set/alpha-8            35.0ns ± 3%    34.1ns ± 2%   -2.43%  (p=0.017 n=10+10)
Set/alpha16-8          36.3ns ± 4%    35.8ns ± 5%     ~     (p=0.254 n=10+10)
Set/gray-8             19.8ns ± 1%    19.7ns ± 0%   -0.69%  (p=0.002 n=8+6)
Set/gray16-8           36.0ns ± 1%    36.4ns ± 2%   +1.08%  (p=0.037 n=10+10)
Set/paletted-8         39.1ns ± 0%    39.6ns ± 1%   +1.30%  (p=0.000 n=10+10)
RGBAAt-8               3.72ns ± 1%    3.58ns ± 1%   -3.76%  (p=0.000 n=9+10)
RGBASetRGBA-8          4.35ns ± 1%    3.70ns ± 1%  -14.92%  (p=0.000 n=10+10)
RGBA64At-8             5.08ns ± 1%    3.69ns ± 1%  -27.40%  (p=0.000 n=9+9)
RGBA64SetRGBA64-8      6.65ns ± 2%    3.63ns ± 0%  -45.35%  (p=0.000 n=10+9)
NRGBAAt-8              3.72ns ± 1%    3.59ns ± 1%   -3.55%  (p=0.000 n=10+10)
NRGBASetNRGBA-8        4.05ns ± 0%    3.71ns ± 1%   -8.57%  (p=0.000 n=9+10)
NRGBA64At-8            4.99ns ± 1%    3.69ns ± 0%  -26.07%  (p=0.000 n=10+9)
NRGBA64SetNRGBA64-8    6.66ns ± 1%    3.64ns ± 1%  -45.40%  (p=0.000 n=10+10)
AlphaAt-8              1.44ns ± 1%    1.44ns ± 0%     ~     (p=0.176 n=10+7)
AlphaSetAlpha-8        1.60ns ± 2%    1.56ns ± 0%   -2.62%  (p=0.000 n=10+6)
Alpha16At-8            2.87ns ± 1%    2.92ns ± 2%   +1.67%  (p=0.001 n=10+10)
AlphaSetAlpha16-8      3.26ns ± 1%    3.35ns ± 1%   +2.68%  (p=0.012 n=8+3)

Fixes #14884

Change-Id: Ia0383530596a550e1b1c7aafce5220e5e0935a53
Reviewed-on: https://go-review.googlesource.com/137495
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>