bench: reduce memory variance in cp and numfmt benchmarks

bench: reduce memory variance in cp and numfmt benchmarks #187