Would it make sense if we add a schema checker/fixer that manually changes uint to int and any timestamp precision to US before writing, so users and other libraries don't pass incorrect schemas during writing.