gh-156347: Fix the documentation of curses window.encoding

The attribute is not used to encode str arguments on a build linked
against a wide-character version of the curses library, which is the
default: str is passed to curses as wide characters and the encoding of
the current locale determines the bytes.  It is used on a narrow build,
where it also decodes results.

Author: fedonman

Doc/howto/curses.rst

@@ -298,8 +298,7 @@ the next subsection.
 The :meth:`~curses.window.addstr` method takes a Python string or
 bytestring as the value to be displayed.  The contents of bytestrings
 are sent to the terminal as-is.  Strings are encoded to bytes using
-the value of the window's :attr:`~window.encoding` attribute; this defaults to
-the default system encoding as returned by :func:`locale.getencoding`.
+the encoding of the current locale (see :func:`locale.getencoding`).
 
 The :meth:`~curses.window.addch` methods take a character, which can be
 either a string of length 1, a bytestring of length 1, or an integer.

Doc/library/curses.rst

@@ -2002,7 +2002,11 @@ Other
 
 .. attribute:: window.encoding
 
-   Encoding used to encode method arguments (Unicode strings and characters).
+   Encoding used to encode and decode method arguments and results
+   (Unicode strings and characters) on a narrow build.  It is not used on a
+   build linked against a wide-character version of the underlying curses
+   library, where strings are passed to curses as wide characters and the
+   encoding of the current locale determines the bytes.
    The encoding attribute is inherited from the parent window when a subwindow
    is created, for example with :meth:`window.subwin`.
    By default, current locale encoding is used (see :func:`locale.getencoding`).