Tooling request: use a large test corpus

[SKalt]

We could run the resulting parser over all the git commit messages from a long-running OSS project such as the Linux kernel to identify parser bugs and performance issues. This parser shouldn't detect any errors since the commit messages are (by construction) valid.

Would you be interested in a PR with a script that could run this kind of fuzz/benchmarking test?

[the-mikedavis]

The commit messages themselves almost never have parse failures - errors tend to come from trying to parse the description in the comments. Those comments are dropped when performing the commit so we're left with just basic messages which are very easy to parse.

Tree-sitter grammar repositories for some programming languages have integration tests that clone some popular libraries and run the parser in test mode against those. I don't think it makes sense to do that for commit messages though since there aren't any really good examples with comments. It would be good to expand the unit tests though with more cases