Detect self equivocation and self-censor

[Stebalien]

If an SP runs multiple lotus nodes, these nodes may end up equivocating. We should be able to detect this pretty quickly in F3 and censor ourselves. We probably want:

1. GPBFT to indicate that it has seen an equivocating message from some specific participant.
2. The client integration to watch this stream for keys it's responsible for. If it sees one, we should either bow out entirely or detect if we should continue/stop based on our peer ID? Maybe?

[Kubuxu]

WAL will facilitate the knowledge of the messages we send. It could trigger tracking of IDs for which we participate. I'm not sure what the strategy should be when we detect it, but I will think about it.

[Stebalien]

I don't think we need WAL for this. We know the identities we're signing for.

[Stebalien]

Discussed sync:

1. We may not catch all equivocations in GPBFT itself, need to check this.
2. We can't just use the WAL + pubsub because we won't propagate equivocations(?).

Hm. The fact that we don't propagate them is actually a bit of an issue... really, the best answer here is to find a better way to avoid equivocations in the first place.

[Stebalien]

Kind of a dumb solution but we could use pubsub for this by broadcasting our leases (likely on a different channel). That would "solve" this issue, albeit at the cost of sending a message to literally everyone (not a huge deal, TBH).

[Kubuxu]

It also reveals the pubsub <-> minerID mapping which is not that great