Get local docker network running again.

[shawn-zil]

Took a while to debug this issue - but I observed a lot of TCP retransmissions occurring in the present local docker network using wireshark, which hinted at an underlying networking issue.

Turns out that the nodes were communicating with each other over the gateway acting as a NAT-router. So, they were seeing the gateway IP/port as the remote IP/port. This is fine if it was used for responses (aka NAT) but would have problems if a client tried to initiate a connection to this ephemeral IP/port that isn't listening on the gateway.

This PR does two things:

• It makes node3 a dependency for all other nodes, as it is the bootstrap node.
• It manually adds the remote IP/ports to the list of known peers exploiting Identify.
• It manually adds the external IP/port (as observed from the remote node) to the list of external addresses.

Local docker cluster is starting up again.

I will be working on my #1878 changes on top of this branch. Would be good to get this fix into main.

[github-actions]

Bencher Report

Branch	2028-docker-network-misconfig
Testbed	self-hosted

Click to view all benchmark results

Benchmark	Latency	Benchmark Result nanoseconds (ns) (Result Δ%)	Upper Boundary nanoseconds (ns) (Limit %)
process-empty/process-empty	📈 view plot 🚷 view threshold	8,312,800.00 (-9.00%)	10,432,499.46 (79.68%)
produce-full/produce-full	📈 view plot 🚷 view threshold	1,739,700,000.00 (-16.98%)	2,803,994,289.61 (62.04%)

🐰 View full continuous benchmarking report in Bencher

[shawn-zil]

Just stashing it here that Rust's libp2p implementation implies that Kademlia needs to be hooked up to Identify, manually, which is largely what got added here.
https://docs.rs/libp2p-kad/latest/libp2p_kad/#important-discrepancies