Schemaless can be accomplished with well-formed formats like json, xml, yaml, toml, etc. from the producer side these are roughly equivalent interfaces. There's zero upside to using CSVs except to comfort your customer. Or maybe you have centered importing of CSVs into your actual business, in which case you should probably not exist.
Until I worked quite a bit with them and realized that there's always schema in the data, otherwise it's just random noise. The question is who maintains the schema, you or a dbms.
Re. formats -- the usefulness comes from features (like format enforcing). E.g. you may skip .ini at all and just go with lines on text files, but somewhere you still need to convert those lines to your data, there's no way around it, the question is who's going to do that (and report sane error messages).