Support for not-a-number or non-finite floating point values

[HolgerPeters]

As I read the edn spec, it does not provide functionality for NaNs or other non-finite floats. This is a drawback already in formats such as JSON. I think there should be a way to fully en- and decode valid floating point values of the IEEE standard, otherwise people would need to refer to using nil or other good old times ways of encoding these (like -9999, etc.).

[willglynn]

This is a dupe of #2.