Investigate resiliance of hydra when one node goes offline and a tx is submitted

[noonio]

We have a test that demonstrates that if a peer goes offline, and a tx is submitted, the head gets stuck (no further snapshots are signed).

We further have a PR that shows that the raft spike fixes the the issue.

Related issues

• Heads stuck in a state without being able to progress snapshots #1773

Solution ideas

• Full raft implemenation
• "Always be in a recoverable state" - Clear pending transactions?
  • Here we'd need people to also go back to the previous snapshot, as the example above demonstrates that 2 out of 3 peers did see the snapshot signed.