Do not start from Iteration 0 when the node restarts

[fed-franz]

Summary

Currently, when the node restarts, it restarts consensus from iteration 0. However, this behavior leads to unwanted consequences, as the provisioner is most likely out of sync wrt the network.

Possible Solution

Upon restart, the node should first synchronize its blockchain and, if its tip is synced, wait for the current round to end before start participating in the consensus.

Alternatively, the node could "listen" for consensus messages, infer the current iteration, and "join" at that or the following one.

[autholykos]

This is important in the context of hard-slashing and faults collection. The problem of restarting from iteration 0 is linked to the propagation of different votes (before the restart and after the restart; in the latter case the vote would most likely be that of a timeout).

The proper way to tackle this issue would be to synchronize after a restart, rather than jump on running consensus immediately.