Float literals for NaN, Inf and -Inf

[dlech]

The compiler currently doesn't have special float literals for the values NaN, Inf and -Inf.

DATAF_NAN is currently defined as 0x7FC00000 but this will cause an error when type checking is implemented.