Reconcile table ownership semantics

[molpopgen]

We currently have two representations of each low-level table -- one that owns its data and one that does not.
This design is arguably confusing and arose primarily over concerns about memory safety.
However, it should be fixed to give a clearer API.

We should be able to write something like:

pub struct EdgeTable(TableManager<tsk_edge_table_t>)

where the manager type is responsible for ownership.
Then, EdgeTable is just a smart pointer and can Deref/DerefMut safely to the manager's public API.

[molpopgen]

To prevent any immediate API breakage, we can supply type aliases for the owning table types.
Marking the aliases as deprecated will cause client code lints to pick up on the change.

[molpopgen]

This is not API breaking after all! The OwningXTable can be a type alias for XTable. This alias will be deprecated in 0.13.0 and removed "eventually".

[molpopgen]

#448 managed to do this w/o appealing to any Deref/DerefMut!