[physfs] Serializing floats

[Ryan C. Gordon]

> So could we have a PHYSFS_writefloat32() function or
> is it a reasonable assumption (I suspect not) that float
> endianness matches int endianness and we can fake it
> by passing a type-punned float to PHYSFS_writeSLE32()?
> (**ick**)

Ultimately, that's what a PHYSFS_*float32() would do beneath the 
hood...the bits mean something different, but they get swapped the same way.

--ryan.