* Downcasting from Py_UCS4 to Py_UNICODE is wrong is Py_UNICODE is 16-bit
   wchar_t
 * Remove old special case in getstring(), unicode is now handled separetely