IPLD Resolver: Integration Test

[aakoshh]

Part of #475

Start an in-memory cluster of #465 and run tests on them to show that they discover each other, learn the membership table, and are able to resolve CIDs. Run scenarios for late joiners to see how Gossipsub behaves.

[aakoshh]

I found these tests for reference in libp2p:

• https://github.com/libp2p/rust-libp2p/blob/v0.50.0/protocols/gossipsub/tests/smoke.rs#L143-L181
• https://github.com/libp2p/rust-libp2p/blob/v0.50.0/protocols/ping/tests/ping.rs#L50-L57
• https://github.com/libp2p/rust-libp2p/blob/v0.50.0/misc/multistream-select/tests/transport.rs#L37-L67

They all do a different combination:

• gossipsub: uses MemoryTransport with Swarm::without_executor, manually polling all created Swarms from the test
• ping: uses tcp::async_io::Transport with Swarm::with_async_std_executor, running the created Swarms in their own Task
• transport: uses MemoryTransport with Swarm::with_async_std_executor

In my case, the Service::new sets up its own libp2p::tcp::tokio::Transport and Swarm::with_tokio_executor, then does the polling in Service::run. I suppose I can factor out transport creation to try to use MemoryTransport, but breaking out run would be awkward, so out of these the 3rd example seems to be where we want to be.