`SagaDagIterator` could/should implement `ExactSizeIterator`

[smklein]

Context: https://github.com/oxidecomputer/omicron/pull/3895#discussion_r1297738612

Current Implementation: https://github.com/oxidecomputer/steno/blob/77d8f2003cb26578eaf80dd2982eda64e2abb80d/src/dag.rs#L387-L405

Advantages of doing this:

• It would be easier to iterate over the nodes under saga idempotency tests

[smklein]

After taking a brief look at this, I don't think this is trivial -- though I'd love to reach into SagaDag.graph and just query for the node_count (https://docs.rs/petgraph/latest/petgraph/graph/struct.Graph.html#method.node_count) , I don't think it's so simple.

• The iterator walks over all named nodes, not all nodes.
• The Start, End, and SubSagaStart nodes don't have names, and would thus need to be omitted from the count.

Therefore, we would need to track the count of all subsagas to be able to infer the named node count (graph.node_count() - (2 + subsaga_count)). This would introduce additional accounting that would need to exist test-only cases.