unicode: improve SimpleFold performance by 2x for non-foldable code points

unicode: improve SimpleFold performance by 2x for non-foldable code points

Change SimpleFold to search the CaseRanges table only once when no
folding is specified for the rune (previously up to two searches could
be performed). This improves performance by 2x for runes that have no
folds or are already upper case. As a side effect this improves the
performance of To by roughly ~15%

goos: darwin
goarch: arm64
pkg: unicode
cpu: Apple M1 Max
                     │ base.10.txt  │             new.10.txt              │
                     │    sec/op    │   sec/op     vs base                │
ToUpper-10             11.860n ± 1%   9.731n ± 1%  -17.95% (p=0.000 n=10)
ToLower-10              12.31n ± 1%   10.34n ± 1%  -16.00% (p=0.000 n=10)
SimpleFold/Upper-10     19.16n ± 0%   15.98n ± 1%  -16.64% (p=0.000 n=10)
SimpleFold/Lower-10     32.41n ± 1%   17.09n ± 1%  -47.27% (p=0.000 n=10)
SimpleFold/Fold-10      8.884n ± 4%   8.856n ± 8%        ~ (p=0.700 n=10)
SimpleFold/NoFold-10    30.87n ± 0%   15.49n ± 3%  -49.84% (p=0.000 n=10)
geomean                 17.09n        12.47n       -26.99%

Change-Id: I6e5c7554106842955aadeef7b266c4c7944d3a97
Reviewed-on: https://go-review.googlesource.com/c/go/+/454958
Reviewed-by: Ian Lance Taylor <iant@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Ian Lance Taylor <iant@google.com>