models: fix Bayesian traversal model and test

[aschampion]

• Fix Bayesian traversal choosing entirely the wrong edges as seed edges (!)
• Fix Bayesian traversal summary values for nodes whose activation CMF never reaches 1.
• Do not include disconnected nodes in Bayesian traversal summary.
• Make Bayesian traversal consistent with the sampling model.
• Test all of the above.

[schlegelp]

Thanks Andrew! Are you happy to have this merged?

@sdorkenw FYI

[aschampion]

Yup.