So I was trying to figure out the best way to go about updating the IVM code given the core types flowing through it are changing.
Rather than update the code and have to track down a million type errors and try to change everything all at once it strikes me as being easier to create an `ivm-2` directory and re-implement (potentially using copy-pasta from the ivm dir) the operators, sources and views.

Once `ivm-2` is done we can swap it into the rest of the machinery (ast, query builder, query api) and delete the old `ivm` directory.

This PR lays out the scaffolding for IVM and re-introduces (while also fixing, streamlining and renaming):
- Materialite
- Source
- DifferenceStream

And implements the `NestedIterable`. `NestedIterable` has been renamed to `IterableTree` here. I think `IterableTree` is a bit more accurate of a description since the sibling iterables can exist.

Now that we have the scaffolding and `IterableTree`, plugging in the operators should be relatively straightforward.

Next steps:
- Further define `pull`
- A view implementation
- filter/join/topk
- Wiring together with an AST that supports subqueries / implementing a PipelineBuilder that understands subqueries