feat: Add resilience in case the cluster falls bellow 1 node

[narvikd]

Even if it isn't supported to go bellow 1 leader and 1 node, it can happen.
In that case the node/cluster will be stuck without observables to prevent serious side effects until it reaches the state where the two nodes are again back online.

Maybe that state never comes, that's why it should try to join an existent leader if there's already one.

[narvikd]

Merging this PR can cause the following:

When a node is stuck in a Candidate state, it will trigger elections, which will cause the system to increment the term.
This can lead to a situation where an offline node has the most current state of the system, and the sick node doesn't.
The offline node will be forced to join this node when it comes online. Losing information in the process.

[narvikd]

If a leader is found, the only right option to prevent rewrites it's a reinstall:
hashicorp/raft#530
hashicorp/raft#477
Not relevant but interesting: hashicorp/raft#525

[narvikd]

It seems to work, but this is very dangerous:
In case of a lack of nodes, because the whole cluster crashed it should be reviewed manually to prevent data losses.

[narvikd]

Will not fix.