Unpatch: non-Latin alphabets

[Peter Stephenson <pws@pwstephenson.fsnet.co.uk>]

Since the storage for insert-composed-char after loading is fairly
compact, I added a few well-known alphabets related to Latin to
it.  (The derivation of alphabets is quite different from that of
languages; I think the Greek and therefore Latin and Cyrillic alphabets
derive from Phoenician, now defunct, which is related to Hebrew and
Arabic, although the last three languages aren't Indo-European.  But you
didn't need to know that.  What I'm really trying to say is I haven't
attempted any Asian alphabets, at least depending how you define Asia.
You know what I mean.)

I also split out the function that defines the associative array into
define-composed-chars, which reduces the memory usage of
insert-composed-char itself quite considerably.

There didn't seem any point in posting a rather dull patch for this, so
I'll commit it directly.

-- 
Peter Stephenson <pws@pwstephenson.fsnet.co.uk>
Work: pws@csr.com
Web: http://www.pwstephenson.fsnet.co.uk