For very small blocks significant time is spent in checkSum function,
adding necessary padding. Instead of writing it byte by byte, copy
encoding/binary PutUint functions, which are compiled into single mov.

name            old time/op    new time/op    delta
Hash8Bytes-6       344ns ± 0%     310ns ± 0%   -9.78%  (p=0.000 n=10+9)
Hash320Bytes-6    1.28µs ± 0%    1.25µs ± 0%   -2.58%  (p=0.000 n=10+10)
Hash1K-6          2.51µs ± 0%    2.47µs ± 0%   -1.67%  (p=0.000 n=10+10)
Hash8K-6          15.8µs ± 0%    15.7µs ± 1%   -0.21%  (p=0.023 n=10+10)

name            old speed      new speed      delta
Hash8Bytes-6    23.2MB/s ± 0%  25.7MB/s ± 0%  +10.77%  (p=0.000 n=10+9)
Hash320Bytes-6   249MB/s ± 0%   256MB/s ± 0%   +2.65%  (p=0.000 n=10+10)
Hash1K-6         408MB/s ± 0%   414MB/s ± 0%   +1.70%  (p=0.000 n=10+10)

Change-Id: I3975ee929465c7dd137d0ca757ad3792a004e1a3
Reviewed-on: https://go-review.googlesource.com/54391
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Filippo Valsorda <hi@filippo.io>