Exercise ICS20 timeouts end-to-end against a Cosmos-SDK chain.

[hdevalence]

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

In advance of getting automatic testing as part of ongoing InterchainTest integration, we should manually exercise the end-to-end flow of timeouts against a Cosmos-SDK chain.

This will require a counterparty chain with ICS23 non-inclusion proof support (#3045) and using our Hermes fork with the config option added in penumbra-zone/hermes#3 to create a client with the correct proof spec.

Currently, this is blocked by:

• jmt: tree traversal fails when fetching an unknown key #3082

[conorsch]

To test this, we can make a packet with near-immediate timeouts. This test should be manual, so we can observe behavior and document results.

[conorsch]

@avahowell is going to tackle this near-term. This takes priority over other ibc-related work like the txp additions described in #3108

[conorsch]

Achieved! @avahowell managed to get a local testing setup together to validate the behavior. She's going to write up those testing steps in this ticket, then close out.

[avahowell]

I was able to successfully perform a timeout of a packet from a local osmosis testnet (built using the osmosis make localnet-* tooling) to a penumbra testnet.

The steps were as follows,

1. perform an ibc handshake using hermes
2. perform a transfer using rly, with a timeout height of +5 blocks
3. wait 5 blocks, then run hermes start
4. observe that hermes constructs a timeout, using the nonexistence proof from the penumbra chain, commits it to the osmosis chain,
5. try another transfer and verify that it still succeeds after timeout

A small change to our query api was required. When that's merged, I'll close this issue

#3163

[conorsch]

Closing as completed.

[conorsch]

Let's try this again, but this time using public testnets, not local nets. This is recently made possible by the dep changes in Osmosis testnet.

[aubrika]

This has been demonstrated, per @avahowell