I am tired of rewriting this over and over to cover edge cases where other language standard csv implementations have assertions on the format and formatting I can cannot guarantee are valid. I've written variations that cover far less concerns over the years and figured I'll make a superset of one that does everything I need. I'll likely refactor this at the end again to optimize for more speed or concurrency.
It's not 1.0 so don't use it in a production context until I lock it and get 100% test coverage.