unicode/utf8: remove some bounds checks from DecodeRune

The compiler couldn't quite see that reading p[2] and p[3] was safe.
This change provides a few hints to help it.
First, make sz an int throughout, rather than just when checking the input length.
Second, use <= instead of == in later comparisons.

name                  old time/op  new time/op  delta
DecodeASCIIRune-8     2.62ns ± 3%  2.60ns ± 5%     ~     (p=0.126 n=18+19)
DecodeJapaneseRune-8  4.46ns ±10%  4.01ns ± 5%  -10.00%  (p=0.000 n=19+20)

Change-Id: I2f78a17e38156fbf8b0f5dd6c07c20d6a47e9209
Reviewed-on: https://go-review.googlesource.com/c/go/+/173662


Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>