Suggestion

We really should have tooling which lets us perf test our tooling.
Specifically we'd want to the test to monitor two main metrics:

1. total runtime
2. peak memory usage

There's a few ideas we could use as tests:

1. parse all fixtures to test TS parse + conversion runtime
2. lint our codebase to test type-aware performance
3. lint some predefined set of OSS codebases at specific commits to validate real-world performance (type-aware and non-type-aware)

Maybe we don't need to do this on every PR - perhaps something we could trigger via a bot command (similar to what the TS team does in the TS repo).