runtime,sync: using fastrandn instead of modulo reduction

runtime,sync: using fastrandn instead of modulo reduction

fastrandn is ~50% faster than fastrand() % n.
`ack -v 'fastrand\(\)\s?\%'` finds all modulo on fastrand()

name              old time/op  new time/op  delta
Fastrandn/2       2.86ns ± 0%  1.59ns ± 0%  -44.35%  (p=0.000 n=9+10)
Fastrandn/3       2.87ns ± 1%  1.59ns ± 0%  -44.41%  (p=0.000 n=10+9)
Fastrandn/4       2.87ns ± 1%  1.58ns ± 1%  -45.10%  (p=0.000 n=10+10)
Fastrandn/5       2.86ns ± 1%  1.58ns ± 1%  -44.84%  (p=0.000 n=10+10)

Change-Id: Ic91f5ca9b9e3b65127bc34792b62fd64fbd13b5c
Reviewed-on: https://go-review.googlesource.com/c/go/+/353269
Trust: Meng Zhuo <mzh@golangcn.org>
Run-TryBot: Meng Zhuo <mzh@golangcn.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>