Mismatch Nim Documentation Regarding 32-bit Floating-Point Literals.

[Araq]

The intention here is to allow `0xaaa'f32` which is a float32 number of the bit 
pattern 0xaaa. It's possible and that both the grammar and the implementation 
are wrong but it suffices to use the issue tracker for it as there is no 
controversial design aspect here. I hope.