Add a divan-based benchmark suite (benches/grep_bench.rs) modeled on the
sed benchmarks, covering literal/fixed-string/regex search, case-insensitive
matching, counting, inversion, line numbers, word/whole-line matching, and
quiet mode. Wire up the dev-dependencies (codspeed-divan-compat, tempfile,
uucore benchmark feature) and a [[bench]] entry.
Add a Benchmarks GitHub Actions workflow running the suite through CodSpeed,
adapted from the sed setup.
