Storage: Use the same tools for short and long term serialization

[liamhuber]

Right now short term serialization (for executors) uses __get/setstate__ and handles everything, while long term (file storage) serialization uses a bespoke solution and only handles nodes that live in a .py file.

Since we'd like so support (with a warning maybe) long-term storage of in-notebook nodes, we'll need to bring in some element of the pickle-style storage in there anyhow. Thus, there is probably room to get these two styles of serialization using the same infrastructure -- namely __get/setstate__. This was my initial attack prior to #160, and has since been requested by @jan-janssen, so it's worth taking another crack at this approach.

[liamhuber]

Everything except for the while-loop is now fully pickle-compliant (as long as you don't stick anything un-pickleable as data on the node; #293 #303 etc), and as long as all the nodes are somewhere importable (i.e. not in main), you can even use the standard concurrent.futures.ProcessPoolExecutor (#304).

So this is actually just waiting on a pickle-compliant long-term storage solution and I will close it.