"cold" exit node startup and missing cfg ref

[sarnold]

Where "cold" startup means the node has not (recently) orbited the root; this issue has not been seen on a node restart after missing the initial cfg ref. If it happens, the root never gets the cfg msg from the ctlr node, because the ctlr node never sees the node msg, so the staging queue is never filled, so there's no bootstrap (ie, no networks are created or configured).

• mbr node: gets the first msg ref, but fails to get the cfg msg (so keeps trying)
• root node: sees the cfg request but never responds
• ctlr node: never sees the pub msg

[sarnold]

The "timing" issue seems mitigated by some refactoring in the announce message queue handling (we now allow duplicate nodes, still only if the node ID and msg ID match).

[sarnold]

Commit f594a44 has all the mitigation stuff for now; updates will be added here if seen again.

[sarnold]

* Where "mitigation stuff" includes the hold queue and a re-insertion of the node ID into the reg queue if hold limit is exceeded.