TODO: Add states/nodes to MarkovChain/DiGraph

[oyamad]

Todo remainder, following the discussion in the Julia side:

• Add states to MarkovChain
• Add nodes to DiGraph

Cf.

• networkx.Graph.add_node
• What to use as nodes and edges

[jstac]

A reminder: Once states are added to MCs on the Python side we should do the same on the Julia side.

[albop]

One quick thought: it may not be efficient to create an set of states 0:n-1 when the markov change is initialized. Knowing that we have the canonical states may allow to speed up some algorithm (like simulation). In that case, it would be better to either keep None for the list of states or to register somewhere that we have the canonical nodes with a dedicated field (e.g. canonical_states in (true, false) )
Another thing: what is the right format for the states ? In my opinion, it should be either None, or a 2d matrix, where each line is a point in a multi-dimensional space. This would make the type quite predictible and cover most applications.

[jstac]

@albop Good point about efficiency.

Aside: It's frustrating that we have these algorithm-level discussion in the py or jl issue trackers and end up duplicating comments --- in this case with QuantEcon/QuantEcon.jl#22.

Regarding the right format for the states, in the majority of cases people will want to work with an array of scalars. Do we need anything more sophisticated?