Having a good quality and curated regex benchmarks would be quite useful! I hope you plan on having several features, and engines that can only have partial support. That would make for very interesting comparisons.
It does. And more. The only thing you have to do is provide a short program that parses the description of the benchmark on stdin, and then output a list of samples that consist of the time it took to run a single iteration and the "result" of the benchmark for verification. The harness takes over from there. There's no need to have any Unicode support at all. I even have a program for benchmarking `memmem`, which is of course not a regex engine at all.