[consensus] Reconsider Performance Penalty on Empty Response

[patrick-ogrady]

If a peer doesn't reply with anything "useful", we don't adjust their performance score. This opens the door to an attack where some malicious peer is good and then only sends empty responses before the timeout (meaning that the node always makes a handful of requests to people that'll never give them something useful first).

We should instead penalize a peer for not giving us something interesting. We should only do this if the response is "empty" and not a response of data we no longer need (but asked for).

Related:

• monorepo/consensus/src/simplex/actors/resolver/actor.rs

  Lines 586 to 600 in 1656a04

  	// Update performance
  	let mut shuffle = false;
  	if !notarizations_found.is_empty() || !nullifications_found.is_empty() {
  	self.requester.resolve(request);
  	debug!(
  	sender = ?s,
  	notarizations = ?notarizations_found,
  	nullifications = ?nullifications_found,
  	"response useful",
  	);
  	} else {
  	// We don't reward a peer for sending us a response that doesn't help us
  	shuffle = true;
  	debug!(sender = ?s, "response not useful");
  	}

• monorepo/consensus/src/threshold_simplex/actors/resolver/actor.rs

  Lines 588 to 602 in 1656a04

  	// Update performance
  	let mut shuffle = false;
  	if !notarizations_found.is_empty() || !nullifications_found.is_empty() {
  	self.requester.resolve(request);
  	debug!(
  	sender = ?s,
  	notarizations = ?notarizations_found,
  	nullifications = ?nullifications_found,
  	"response useful",
  	);
  	} else {
  	// We don't reward a peer for sending us a response that doesn't help us
  	shuffle = true;
  	debug!(sender = ?s, "response not useful");
  	}