PoS/test: division by zero

[tzemanovic]

Trace: https://github.com/anoma/namada/actions/runs/5253847650/jobs/9491877974?pr=1549

The state machine test failure is most-like a problem with pre-conditions because get_total_consensus_stake ends up being zero.

[brentstone]

I have a local change that somewhat fixes this issue actually. I started doing SM tests with somewhere around 8 consensus validator slots and more genesis validators. Compared to the current config of like 3 or 4 consensus validators, we run into this issue MUCH less frequently (i.e. I never have yet).

At one point I tried to rigorously avoid this issue, but it ended up being tricky, so I abandoned that in favor of this hack.

[brentstone]

@tzemanovic Are we still concerned with this enough to keep the issue open? Haven't encountered this in a while but don't recall if we've explicitly solved it.

[tzemanovic]

we have a fix for this in #1612 where we account for enqueued slashes in the pre-condition (

namada/proof_of_stake/src/tests/state_machine.rs

Lines 2657 to 2668 in d0f994f

	// Find the number of enqueued slashes to unique validators
	let num_of_enquequed_slashes = state
	.enqueued_slashes
	.iter()
	// find all validators with any enqueued slashes
	.fold(BTreeSet::new(), |mut acc, (&epoch, slashes)| {
	if epoch > current_epoch {
	acc.extend(slashes.keys().cloned());
	}
	acc
	})
	.len();

)