Avoid another copy in RE2::PossibleMatchRange, and simplify the code.

Again, this improves the case when the prefix is a large proportion of
maxlen. Gains are ~20% in this case.

Before:

Benchmark                    Time(ns)        CPU(ns)     Iterations
-------------------------------------------------------------------
PossibleMatchRange_Trivial       1158           1156         513274
PossibleMatchRange_Complex       2593           2589         272939
PossibleMatchRange_Prefix          71.1           71.0      9838330
PossibleMatchRange_NoProg          71.3           71.2      9844569

After:
Benchmark                    Time(ns)        CPU(ns)     Iterations
-------------------------------------------------------------------
PossibleMatchRange_Trivial       1134           1132         617832
PossibleMatchRange_Complex       2545           2542         276152
PossibleMatchRange_Prefix          56.5           56.4     10000000
PossibleMatchRange_NoProg          57.7           57.6     10000000

Change-Id: Ie3c6c4ea0b4de3575675fbca4856734a5f5120cc
Reviewed-on: https://code-review.googlesource.com/17730
Reviewed-by: Paul Wankadia <junyer@google.com>
4 files changed