Add more functionality to floating point number type

Currently, the only way to create floating point numbers is to parse them from string. Pretty much the only useful things I can do with these floating point numbers is arithmetic (addition, multiplication, subtraction and division), as well as turning them back into strings. In this sense they're a finite precision alternative to the built-in rational type, albeit with less operations.

I propose adding the following (at the very least):
- Double.exp and Double.log
- Double.floor (and/or truncate, or something)
- Double.toInt

Here is a related issue: https://github.com/CakeML/cakeml/issues/824.