[haskell-mode] Work around coding cookie breakage

[Stephen J. Turnbull]

Joachim Schrod writes:

 > AFAIU, UTF-8 support is only available in Mule,

I don't think that's true (it's not true on Windows, but I'm not sure
about *nix).  In any case, adding UTF-8 support to 21.5 no-Mule is not
very difficult.

 > as long as the Great Rewrite To Internal Unicode has not
 > happened. And no future release of XEmacs should go out without
 > UTF-8 support.

All currently maintained XEmacsen do support UTF-8 (albeit with the
help of Mule-UCS in the case of 21.4).  However, there remain lots of
people, at least in the U.S., who don't need anything beyond
ISO-8859-X, and who do need efficient buffer addressing, which Mule
cannot currently offer.  The option to disable Mule is therefore
desirable.

 > Therefore, I would have thought that getting rid of no-Mule is
 > unavoidable.

In the long run, yes, but I think we need widechar support for that.
According to Ben (and based on the experiences of Samba and Python),
that will be difficult and require lots of beta testing == time.