If running a benchmark with --ignore-failure, it might be interesting to know if some of the commands actually failed.

For example, consider a long running compile-time benchmark:

hyperfine \
  --runs 3 \
  --ignore-failure \
  --parameter-list git_tag "v0.1,v0.2,v0.3,v0.4,v0.4.1,v0.5" \
  --prepare "git checkout --force {git_tag}; make clean" \
  --export-json "compile_time.json" \
  "make -j8"

We pass --ignore-failure because we don't want all benchmark results to be lost if one of the tags does not compile cleanly. However, at the end of the run, we would like to know if one of the commands actually failed.

We could add a exit_codes: Vec<i32> field to BenchmarkResult such that the exit codes would be exported to the JSON file.