Support for 32bit and 64bit floats

[cristian-5]

Add support to convert a float literal to its binary representation, effectively adding the types f32 and f64. Note that the float literal could be expressed with exponential notation as well.

[chrisgbk]

This was discussed in #92 a while ago, with some goals and design considerations - namely, a desire not to lock users into IEEE-754 floating point numbers and support a mix of custom floating point and fixed point floating point numbers.