Add tombstoning to smoke test

[hdevalence]

Is your feature request related to a problem? Please describe.

We don't currently have an integration test that checks that a validator who double-signs a block actually gets tombstoned, and since this is kind of annoying to manually run, we've never actually executed that code path. Attempting to fix #1207 revealed that the necessary indexes weren't being populated, so tombstoning wouldn't have worked, even if it had happened. This is bad, and exposes a gap in our testing.

Describe the solution you'd like

We should add a tombstoning test to the smoke test. This would involve (1) creating a new validator (2) inducing it to double-sign (3) checking that it is tombstoned. One way to induce double-signing is to run two instances of tendermint+pd that have a common priv_validator_key.json.

[conorsch]

Given recent progress toward #1664, I assume we'd want to run the tombstoning check in rust code against the state machine, rather than via integration tests. Do you agree, @hdevalence?

[conorsch]

Scratch that: we discussed at sprint-planning today, and having this as a true integration test is a Good Thing:tm:. Reducing priority for now, though.

[conorsch]

I'll take a first look at this for 44; might have to reorganize the compose files a bit to make coerce the double-sign action. Not sure it'll ship in 44, but will document progress as I go.

[conorsch]

Took a first pass. I'd like to wire up a new CI workflow much like the smoke test, but that uses two validators, one with a duplicated identity. The small changes suggested in #1992 will make this a bit more straightforward, so I'm going to tackle that first.

[hdevalence]

We won't be adding any new checks to the smoke test. This should be done as a staking state machine test.