This would break format strings like memory://%.*ws (that dashboard tries using), because the "ws" part would set FF_IsWide, but FF_IsWide actually seemed to be used as an "is opposite encoding" flag.
Since the format string is wide that'd make it treat the param as ASCII, but in dashboards case the param string provided is actually a wide string.
(in other words, it seems %ws is supposed to mean that the param is always a wide string, not that the param is just opposite encoding)
This changes the old FF_IsWide to FF_InvertWide, and adds a new FF_IsWide that means that the param is always wide.
9185cdcc79