XIM patches

[Owen Taylor <owt1 cornell edu>]

OK, I've gone ahead and added Takashi Matsuda's XIM patches
to the CVS repository. I'm fairly confidant that they work
pretty well, but they add a lot of code that deals with
complicated things, so there may well be bugs lurking in
the changes despite our best efforts.

I suspect the most likely area of difficulty is portability.
The code will use -DX_LOCALE for locales unsupported by
the C library; if not using -DX_LOCALE, things may be
somewhat at the mercy of the operating systems support for
wide/multibyte character strings. (This is just a guess,
for all I know it works perfectly on every system.)

Configuring with --disable-xim will disable most of the
changes, so that might be a workaround if there are systems
on which this stuff causes major problems. (hopefully not!)

It's been tested with the Japanese input method kinput2
(and maybe others), and seems to work OK with the built
in input methods for the ISO-8559-1 and -2 locales.
(Yes, compose and dead keys now work with this patch.)
I don't think it has been tried out yet for Greek, Cyrillic,
or other Asian locales.

Unfortunately, I'm going to be away until Sunday, so any 
problems with the changes should be reported to this list 
and/or to the author, Takashi Matsuda <matsu@arch.comp.kyutech.ac.jp>.

Regards,
                                        Owen