Standalone 'info' should recode into display's encoding

[Paul Eggert]

If I have an info file containing '@documentencoding UTF-8',
makeinfo generates an Info file encoded in UTF-8.  But then if
I attempt to read the file using the standalone 'info' reader
in (say) the 'en_US.iso885915' locale, 'info' misbehaves and
puts funny characters onto the display.

This is a problem with GNU Emacs, which ships .info files as
part of its distribution.  In the latest stable version of
Emacs (24.3), for example, gnus.texi contains this:

  @samp{»} is translated into @samp{>>}, and so on.

where '»' is the UTF-8 character.  makeinfo 5.2 converts this to:

  ‘»’ is translated into ‘>>’, and so on.

where '»' is still the UTF-8 character and there are
UTF-8 curly single-quotes; so far, so good.
But when I run the standalone 'info' command in the
en_US.iso885915 locale, I see this on the screen:

  â»â is translated into â>>â, and so on.

because 'info' merely sends the bytes of the UTF-8 representation
to the screen.

I'd rather see something like this:

  '»' is translated into '>>', and so on.

where this '»' is the unibyte ISO-8859-15 character '\273',
and the UTF-8 curly single-quotes are transliterated to
ASCII apostrophes.

For my preferred platforms (GNU/Linux based), this is not much
of a problem, since I rarely use unibyte locales.  But when the
topic of formatting manuals containing UTF-8 came up on the
Emacs bug list, a Microsoft Windows user objected that the
manual would look bad on his screen, I suppose because he
uses a unibyte locale.  For more details please see Emacs bug#16292 at:

http://bugs.gnu.org/16292