Fix exponentiation and unary minus precedence in abakus (I haven't looked at this stuff in ages, I'm glad I remembered to comment it.)

Now -2^2 gives a result of -4 as expected. 2^3^2 still gives the right answer, all other expressions I've tested still work. I haven't developed a testsuite yet and I saw no online testsuite for math parsers so this may need more testing.

Noted on a bug report by Vincent Lefevre.

I don't intend to backport to the the KDE 3.5 version of abakus (but because I don't feel like trying to find out where I left it, not because I'm opposed).