Explain in the documentation that conversion errors should never happen.

 * Add error_pos optional argument to _Py_wchar2char()
 * PyUnicode_EncodeFS() raises a UnicodeEncodeError or MemoryError if
   _Py_wchar2char() fails

_Py_char2wchar() callers usually need the result size in characters. Since it's
trivial to compute it in _Py_char2wchar() (O(1) whereas wcslen() is O(n)), add
an option to get it.

surrogate characters.

surrogate characters.

On Windows, Py_UNICODE is wchar_t, so we can avoid the expensive Py_UNICODE*
=> wchar_t* conversion.

 * Don't define _Py_wstat() on Windows, Windows has its own _wstat() function
   with a different API (the stat buffer has another type)
 * Include windows.h

I suppose that I reapplied my local patch creating Python/fileutils.c whereas
the file already existed.

 * _Py_fopen() and _Py_stat() come from Python/import.c
 * (_Py)_wrealpath() comes from Python/sysmodule.c
 * _Py_char2wchar(), _Py_wchar2char() and _Py_wfopen() come from Modules/main.c
 * (_Py)_wstat(), (_Py)_wgetcwd(), _Py_wreadlink() come from Modules/getpath.c