need to recognize non-ASCII identifiers

[Dave Love]

cc-mode needs to recognize non-ASCII identifiers.  I guess this should
be done for all languages, since Java is defined to use Unicode and
you can use non-ASCII charsets in C.

This will only work cleanly in Emacs 21, which should use replacements
in regexp char classes like `[:alnum:]' instead of `a-zA-Z0-9' and
similar in the relevant places.  (It should be safe to make the
replacement generally.)  To make it work also with XEmacs, you can
test for the relevant features with something like

`(string-match "[[:alpha:]]" "x")'

and conditionalize on the result at compile time.

I tried to make suitable changes but then failed to build the result
properly because of the way it's compiled, which I don't understand;
I've commented on that before...  I guess it's easy to start from
scratch on the replacements.