[mo4islona]

Moves sqd-network from 11f4230, a commit of the closed subsquid/sqd-network#216 branch, to e1460fd on main, which has subsquid/sqd-network#217.

Why

11f4230 looks a worker up again after a failed dial, but it stops retrying when a lookup cannot find a worker that is still offline: the next dial fails with NoAddresses, which it does not retry. It also lacks the open_stream hang fixes in the vendored libp2p-stream. The pings collector and the logs collector keep connections to all workers, so both need the full fix.

Changes

• sqd-network e1460fd for all three crates. sqd-messages is now 3.1.0.
• The [patch.crates-io] for the libp2p fork is removed: sqd-network now uses libp2p from crates.io.
• Test-only updates for the new API: DataChunk::new takes &str and validates the hash, and Query has a new chunk_version field. The collectors do not store that field, and query signatures do not cover it.
• The router strips /blocks.parquet from an object key before it parses the chunk id. The shared DataChunk parser reads everything after the second dash as the hash, so it rejects the full key, which the old parser accepted.
• pings-collector 2.9.1, logs-collector 2.5.1, portal-logs-collector 1.1.7, so the release does not overwrite published tags.

Testing

cargo fmt --all -- --check, cargo clippy --workspace --all-targets --locked -- -D warnings and cargo test --workspace --locked pass.

🤖 Generated with Claude Code