Use subslices with known length and cap to give bounds checking hints
to the compiler. Improves over the earlier pointer based optimizations
in https://go-review.googlesource.com/c/go/+/14093 for GlyphOver but
not for FillOver so the latter is left unchanged.

See #27857 for discussion of small caps used in subslices.

name               old time/op  new time/op  delta
FillOver-8          607µs ± 1%   609µs ± 1%     ~     (p=0.447 n=9+10)
FillSrc-8          23.0µs ± 1%  22.9µs ± 2%     ~     (p=0.412 n=9+10)
CopyOver-8          647µs ± 0%   560µs ± 0%  -13.43%  (p=0.000 n=9+10)
CopySrc-8          19.3µs ± 1%  19.1µs ± 2%   -0.66%  (p=0.029 n=10+10)
NRGBAOver-8         697µs ± 1%   651µs ± 1%   -6.64%  (p=0.000 n=10+10)
NRGBASrc-8          405µs ± 1%   347µs ± 0%  -14.23%  (p=0.000 n=10+10)
YCbCr-8             432µs ± 2%   431µs ± 1%     ~     (p=0.764 n=10+9)
Gray-8              164µs ± 1%   139µs ± 1%  -15.44%  (p=0.000 n=10+10)
CMYK-8              498µs ± 0%   461µs ± 0%   -7.49%  (p=0.000 n=10+9)
GlyphOver-8         220µs ± 0%   199µs ± 0%   -9.52%  (p=0.000 n=9+10)
RGBA-8             3.81ms ± 5%  3.79ms ± 5%     ~     (p=0.549 n=9+10)
Paletted-8         1.73ms ± 0%  1.73ms ± 1%     ~     (p=0.278 n=10+9)
GenericOver-8      11.0ms ± 2%  11.0ms ± 1%     ~     (p=0.842 n=9+10)
GenericMaskOver-8  5.29ms ± 1%  5.30ms ± 0%     ~     (p=0.182 n=9+10)
GenericSrc-8       4.24ms ± 1%  4.24ms ± 0%     ~     (p=0.436 n=9+9)
GenericMaskSrc-8   7.89ms ± 1%  7.90ms ± 2%     ~     (p=0.631 n=10+10)

Change-Id: I6fe1b21bb5e255826cbfdd2e73efd5858cd5557c
Reviewed-on: https://go-review.googlesource.com/136935Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>