I'm surprised this doesn't mention the class of parsers I believe to be the most useful - the database in question itself. For example I make extensive use of https://pglast.readthedocs.io/en/v3/, which is the postgres SQL parser extracted into a standalone library (pg_query). No 3rd party library attempting to parse a particular flavour of SQL will be more accurate than the database that defines that flavour, since you're also getting all of the edge cases and quirks of it's impementation.