Fix bug with float literal constant

[averycowan]

There are no float literals. If you attempt to simulate one using a cast and division of integers, the optimizer precomputes the desired float value and tries to represent it. This fails and crashes compilation.

[leijurv]

If you attempt to simulate one using a cast and division of integers, the optimizer precomputes the desired float value and tries to represent it. This fails and crashes compilation.

uh then how does this work?
https://github.com/leijurv/Kitteh2/blob/master/test/tests/mandelbrot.k#L4

[niuggg]

goodproject