Test power-table lookback and message queue means no rebroadcast needed when participants lag

[anorth]

When participants are within $lookback instances of each other, no rebroadcast(or even round skipping) should be necessary for all nodes to get in sync once synchrony is re-established. Test this with multi-instance simulation test, artificial message delays, and metrics on rebroadcasting.

See also #125

[Stebalien]

and metrics on rebroadcasting

Or just disable rebroadcast?

Specifically, this is testing "skip to instance".