This is very similar to the "lazy strings" idea.

unicode_internal codec uses Py_UNICODE instead of the real internal
representation (PEP 393: Py_UCS1, Py_UCS2 or Py_UCS4) for backward
compatibility.

when called with a timeout.  Patch by Arnaud Ysmal.

when called with a timeout.  Patch by Arnaud Ysmal.

getcwdb() => getcwd()

 * unicode_decode_call_errorhandler() now supports the PyUnicode_WCHAR_KIND
   kind
 * unicode_decode_call_errorhandler() calls copy_characters() instead of
   PyUnicode_CopyCharacters()