network: preserve reqresp error codes

[zclawz]

Summary

• bump zig-libp2p to blockblaz/zig-libp2p@a46f4e1
• use the new coded reqresp error API from zeam's inbound server stream
• preserve INVALID_REQUEST response-code bytes for malformed/invalid reqresp requests instead of just closing the stream

Supporting dependency PR: blockblaz/zig-libp2p#300

Hive context

Suite: https://hive.leanroadmap.org/suite.html?suiteid=1784546141-eb3090c7f9b766cd983e0fb631c26ad7.json&suitename=reqresp&client=zeam_devnet5

The zeam-owned reqresp failures include two explicit protocol assertions:

• reqresp/blocks_by_range/zero_count: expected INVALID_REQUEST
• reqresp/blocks_by_range/too_many_blocks: expected INVALID_REQUEST

Zeam's node handler already calls responder.sendError(constants.RPC_ERR_INVALID_REQUEST, ...), but the zig-libp2p QUIC responder path discarded that numeric code and only closed the response stream. Hive's mock reqresp peer expects an actual response chunk whose first byte is 1, so it observed the wrong wire behavior.

This PR wires the error code through zeam -> zig-libp2p -> QUIC response framing, producing a non-success reqresp response-code chunk before stream close.

Notes

The same Hive report also contains zeam failures where the client does not produce/catch up to non-genesis blocks within the harness timeout. I traced those separately through status-driven sync and block serving; they are not fixed by this narrow wire-protocol patch and likely need a follow-up focused on checkpoint-sync/status catch-up behavior.

The report also includes many non-zeam failures caused by other clients failing startup; this PR only changes zeam-owned behavior.

Validation

• zig fmt --check pkgs/network/src/ethlibp2p.zig
• git diff --check
• zig build --fetch --summary none
• zig-libp2p dependency branch: zig build test --summary all (508 passed, 2 skipped)
• zeam zig build test --summary all was started with Zig 0.16.0, but this local container still cannot complete the repo target because rustup is absent (failed to spawn ... rustup: FileNotFound); lightweight Zig package groups reached before the slow proof/STF phase passed.

[zclawz]

Status: ready for review, with scoped coverage.

This PR addresses the reqresp wire-protocol failures where Hive expected INVALID_REQUEST response code 1 for invalid BlocksByRange requests. The zeam handler already selected code 1; zig-libp2p was dropping it and only closing the stream. Supporting dependency PR: blockblaz/zig-libp2p#300

Validation:

• zeam: zig fmt --check pkgs/network/src/ethlibp2p.zig, git diff --check, zig build --fetch --summary none
• zig-libp2p dependency: zig build test --summary all passed (508 passed, 2 skipped)
• zeam full test target still cannot complete in this local image because rustup is missing; lightweight Zig package groups reached before the slow proof/STF phase passed.

Remaining from the Hive report: the zeam did not produce blocks / catch up from status failures appear to be a separate sync/catch-up root cause, not fixed by this PR.

[zclawz]

Done.

Updated PR #1054 to consume zig-libp2p v0.2.90:

• build.zig.zon now uses https://github.com/blockblaz/zig-libp2p/archive/refs/tags/v0.2.90.tar.gz
• package hash updated to zig_libp2p-0.2.90-lil2hHdyKgAfzB0t9KNYv1G-esu4FB0q1R8Wjp2mJVKI
• pushed commit: 2776c09b5163099c858536411f9086fd393cc5e2

Validation run locally:

• zig fetch https://github.com/blockblaz/zig-libp2p/archive/refs/tags/v0.2.90.tar.gz passed and produced the committed hash
• zig build --fetch --summary none passed
• git diff --check passed

I could not run the full build/test suite in this image because PATH has Zig 0.15.2, while the branch/dependencies require Zig 0.16.0. I attempted to fetch the 0.16.0 toolchain side-by-side under /tmp, but the download was too slow for this Telegram turn, so I stopped it after the dependency update was pushed.

[ch4r10t33r]

LGTM