Recognize charset extensions, see #3150.

Recognize charset extensions, see #3150.

With utf-8//TRANSLIT, we internally didn't recognize it as utf-8. This
leads to badly broken behaviour if --without-wc-funcs is used for some
reason. In that case, if we have utf-8 as charset, we implement our own
wide char functions; for all other charsets, we use the system
single-byte locale functions. And using these with utf-8 is broken.