fix(network): bind strict-PQ peer identity to staking ML-DSA key so validators produce blocks

[Darkhorse7stars]

Summary

On a strict-PQ chain a peer's consensus identity is its ML-DSA-65 NodeID (StakingConfig.DeriveNodeID), but the network layer kept every peer on the TLS-cert NodeID derived during the transport upgrade (peer/upgrader.go → ids.NodeIDFromCert). The validator set is keyed by the ML-DSA NodeID, so every peer was classified as a non-validator → the P-chain saw zero connected validators → consensus never formed → no block was ever produced (the built-in EVM/C-Chain stays at height 0, RPC serves reads but eth_blockNumber never advances).

This was observed on the Liquidity strict-PQ devnet: 3 validators, all healthy, all BLS-correct, peer mesh formed — but P-chain height stuck at 0.

Root cause — two coupled defects

1. The PQ handshake signed with an ephemeral key, not the staking key.
network.NewNetwork built the handshake identity with peer.NewLocalIdentity(MyNodeID), which generates a fresh ML-DSA keypair per process (see its doc-comment). So the handshake signature proved possession of a throwaway key with no relationship to the staking key that MyNodeID derives from. The wire carried the right NodeID, but nothing bound it to a key the validator set knows. (Corollary: the handshake never actually authenticated the validator identity — a peer could assert any NodeID.)

2. The verified peer NodeID was discarded.
peer.runPQHandshakeIfRequired used HandshakeResult.AEADKey but dropped HandshakeResult.PeerNodeID, leaving p.id on the transport TLS-cert NodeID. Consensus then looked up that TLS NodeID in an ML-DSA-keyed validator set and found nothing.

Fix

• Thread the node's persistent staking ML-DSA keypair (StakingConfig.StakingMLDSA{,Pub}) onto network.Config (mirrored in node.Node.initNetworking) and build the handshake LocalIdentity from it via the new peer.NewLocalIdentityFromStakingKey. The handshake now signs with the same key that derives MyNodeID.
• peer.adoptVerifiedPQIdentity (new): after a successful handshake, re-derive the NodeID from the peer's presented ML-DSA key under the node-identity domain (ids.Empty — the exact domain DeriveNodeID uses for MyNodeID) and require it to equal the presented NodeID, then adopt that ML-DSA NodeID as p.id.

This both fixes block production and closes an identity-impersonation gap: a peer can no longer claim a NodeID it cannot derive from the key it proved possession of.

Because peer.Start runs the handshake synchronously before the message-pump goroutines and before network.upgrade adds the peer to connectingPeers/connectedPeers, p.id is already the ML-DSA NodeID by the time any peer-set bookkeeping keys by it — no re-keying race.

Blast radius

Entirely inside the strict-PQ path (SecurityProfile != nil && profileRequiresPQHandshake). Classical / permissive chains skip the PQ handshake and are unaffected — p.id stays the TLS-cert NodeID exactly as before. No wire-format change (same INIT/RESP frames); only which key signs, plus an added local verification.

Rollout / review notes

• Coordinated upgrade for strict-PQ networks. The binding check rejects the old ephemeral-key handshake, so all nodes on a strict-PQ chain must run this together.
• Needs a devnet soak (3-validator strict-PQ: confirm P-chain height advances + getCurrentValidators lists all NodeIDs with weight + EVM eth_blockNumber increments) before any production rollout. Drafted for chain-team review; do not blind-merge/deploy to a live network.

Tests

go build ./... clean; go vet clean; go test ./network/peer/... green. Adds network/peer/pq_identity_adopt_test.go covering: bound identity → adopted; unbound/forged NodeID → rejected + identity untouched; nil/empty result → rejected.

Follow-ups (not in this PR)

• The pre-handshake MyNodeID self-check and AllowConnection in network.upgrade still evaluate the TLS-cert NodeID (best-effort pre-filters; authoritative gating is post-handshake on p.id). Worth migrating to the ML-DSA NodeID for completeness.
• The PQ handshake runs under peersLock in network.upgrade; a slow peer serializes connection establishment. Pre-existing; orthogonal to this fix.