Any recommendations for testing frameworks for calling binaries?

 It's probably not exactly what you want, but a longish time ago, I was helping to organize some programming contests. And most of the time, people submitted code, it got compiled and then checked against expected inputs and outputs. It was pluggable, so it could scale from exact output, with some variance in the output (whitespace differences, lines in arbitrary order), to judge programs (ok/not-ok output) to things like interactive opponent games. We were using this for the whole stuff: https://www.ucw.cz/moe/, maybe you could use something from there.