FYI: java.text merge

[Tom Tromey]

I've merged most of java.text between Classpath and libgcj.

I stopped short of merging the collation code.  The two
implementations are different enough that I didn't know what to do.

On the one hand, libgcj tries (it fails due to a bug -- but that is a
separate issue) to implement character decomposition properly.  As far
as I can tell the Classpath implementation doesn't attempt this.  Also
libgcj tries (it has not really been tested) to implement
multi-character collation elements, which I think Classpath does not
(this is needed for Spanish collation, at least).

On the other hand, the Classpath implementation implements
CollationElementIterator.previous().  (The implementation is incorrect, 
due to decomposition not being handled, but that is a separate issue.)
Adding this method to the libgcj implementation is non-trivial.

So, for now, I decided to punt on this problem and leave a few classes
unmerged in java.text.  The current deficiencies in the Classpath
implementation don't really matter since our locale support is so
spotty anyway.  This whole area will need a lot of testing and
debugging whenever full locale support is implemented.

Another issue to figure out is how to implement getAvailableLocales()
in various classes.

Meanwhile, I'm going to check in my merge.  I've built it in the
Classpath tree using jikes.  I've tested this code in the libgcj tree.
I tested it against Mauve, too.  (There are a couple Mauve regressions
from what I remember, but I believe they are actually due to an
earlier change.  Looking at these is on my to-do list.)  Feel free to
ask questions, comment, etc.

Tom