Backported the very useful oracle_compat.c from postgres 8.0 beta to 7.4.5

Dear all!

I'm new to the list, so please excuse if I repost this patch.

In the found the _very_ useful oracle_compat.c function which implements
correct UPPER(), LOWER() and INITCAP() for multibyte charactersets.

This change is very important for everyone who uses multilingual
databases, because it makes case insensitive search possible (like
"Ärzte" and "ärzte"). We have tested the patch in our work, and we did
not see any problems so far. I think Postgres without that patch is buggy.

Thank you to Tom (who wrote the code) for the great work.

The attached patch will patch the current stable version 7.4.5

Best whishes,

Johannes