Need size benchmarks.

@ekrich A **most excellent** start. Bravo!

With reference to your point 4, code size. I suggest that we need a set of
benchmark .scala files, starting small. That is, an executable which uses
Character.toLower, and another which uses regex and a third which uses both.

We could build the executables before any attempt at unifying Character
and re2s and again after, looking for changes in binary size.

I think/suspect that simply measuring the helloWorld Test.scala binary
before and after will not yield useful information.

When I sized the binary for Test.scala before and after my re2s 1 dimensional
array work, I saw no size change. Println probably does not bring in re2s.

@densh had mentioned that some String methods used regex (probably to split or
tokenize).

If/when I dig myself out of this wave of re2s work, I can take a look at creating
a benchmark to show size changes. I may need such for this wave even.

What I probably need is a suggestion for a directory in this repository for both
the source and a historical record of sizes (log file).

Start with the named Test.scala example and locate the repository's existing build or test entry points. Add small benchmark .scala cases covering Character.toLower, regex, and their combination, then build them before and after the relevant change and record comparable executable sizes in the proposed historical log.

Written by the indexing model from the issue text.