#2 minor split() bug

[Kyle Downey]

I ran SpellDictionaryDisk against SCOWL 5
(http://wordlist.sourceforge.net) for all files minus
any larger than .70 and any Canadian and British words.
I ended up with an initial index entry like this

,0,42

that choked SpellDictionaryDisk.loadIndex(); I believe
the split() function isn't handling this case because
of its use of StringTokenizer. Deleting the line let me
work as normal.

Also, a request/question: is there a way to distribute
binary files only with SpellDictionaryDisk, or does it
require the words/*.* files to be distributed to the
users? Just for space/package size reasons it would be
nice to ship just db and idx files.