Ability to Marshal objects with memoized methods

[sferik]

See discussion here: https://github.com/dkubb/memoizable/pull/4#issuecomment-30849404

[mbj]

This is an interesting question. If this would be memoization for adamantium I'd argue we'd not even try to serialize the cache state. As memoizable would also be used for objects with a mutable root state (lets call the memoized stuff child state), I'd say the child state must be serialized also.

I still consider memoization of return values that are calculated from possible mutable state a smell.

@dkubb If we add cache serialization to memoizable we should think about adamantium, and if it should serialize the memoization state, or not.