fix(#123): close all ten devnet RPC parity gaps against local Curio devnet

[Reiers]

Closes #123.

Every change is gated on network == "devnet"; mainnet + calibration paths are byte-identical to pre-change behaviour.

What changes (per finding on #123)

P0 correctness (breaks EVM clients silently):

• [Wire VM bridge to upstream Forest/Lotus for block production state-root computation #4/Embedded web dashboard (served by the daemon at /dashboard) #5] eth_chainId + net_version return the devnet's actual EIP-155 chainId (curio-fork docker: 31415926 / 0x1df5e76) instead of the mainnet fallback 0x13a / "314". devnet-init reads eth_chainId from the running lotus and persists it into devnet-config.json as ethChainID. net_version already delegates to EthChainId inside rpc/server/eth_api.go, so both are fixed by the single handler change.

P1 correctness (breaks gossipsub identity + curio schedulers):

• [Close gossipsub-side sync lag (parent backfill + direct peers) #1] Filecoin.StateNetworkName returns the devnet's localnet-uuid instead of the literal string "devnet". Handler now reads from build.GetDevnetConfig() instead of returning a literal. Aligns /fil/blocks/<name> gossipsub topic + DHT protocol prefix with lotus.
• [Keystore passphrase silently falls back to empty string when LANTERN_PASS is unset #2] Filecoin.Version.BlockDelay returns the devnet's actual cadence (curio-fork: 4) instead of hardcoded 30. devnet-init reads BlockDelay from Filecoin.Version and persists as blockDelaySecs.
• [StateMinerInfo / StateMinerPower timing out at 30s on cold state #3] Filecoin.ChainGetGenesis returns the persisted genesis CID instead of not implemented. Compile-time genesis constants exist only for mainnet + calibration; devnet's genesis is per-boot.

P2 correctness (unblocks send-path + peace-of-mind restart tests):

• [Architecture Questions #10] Filecoin.MpoolPushMessage now works on single-node devnet. The gossipsub mesh can't form on a single-node docker devnet, so the standard mpool.Pool had nothing to publish onto. Daemon now wires a Pool whose Config.Sink POSTs directly to the devnet lotus via Filecoin.MpoolPush. All other Pool semantics (persist journal, pending set, nonce derivation, #47 reconcile/rebroadcast, #121 restart-persist) work identically; only the wire transport changes. Unlocks the entire peace-of-mind restart drill on devnet, which is our fastest iteration loop.
• [JWT tokens are issued without exp claims and never expire #7] eth_getCode / eth_call / eth_getStorageAt cold-state timeouts. Devnet mode auto-wires the devnet lotus as the VMBridge when the operator didn't pass --vm-bridge-rpc. A single-node devnet has no libp2p, so bitswap can't fetch state blocks; the daemon fell through to errBridgeUnconfigured.
• [Beacon cert-exchange upstream should default to other Lantern beacons, not Glif #6] Filecoin.MpoolGetConfig returns a stable snapshot instead of HTTP 500. Values chosen to match lotus defaults. PriorityAddrs is nil (marshals to JSON null, matching lotus's wire shape byte-for-byte); an empty slice would marshal as [] and break consumers that check == null.

Trust posture unchanged from #122: devnet is single-source by design (operator owns both the devnet lotus and the Lantern client).

New surfaces

• mpool.Config.Sink hook lets Pool run with a nil pubsub instance. When Sink is non-nil and ps is nil, New() skips the gossipsub join+subscribe entirely and Publish/Rebroadcast route through the sink. Only used by devnet mode; mainnet/calibration paths continue to require a pubsub instance and reject nil-pubsub construction.
• net/glif.Client.EthChainID, .BlockDelaySecs, .MpoolPush — three new methods on the shared JSON-RPC client. Used by devnet-init (boot-time discovery) and by the daemon (devnet mpool sink).
• build.DevnetConfig gained EthChainID + BlockDelaySecs fields. Older configs written before this change have zero values; handlers hint lantern devnet-init --force to re-capture from the running lotus.

Tests

• 5 new unit tests in net/mpool/sink_test.go cover the nil-pubsub + Sink path: nil-pubsub requires Sink; Publish invokes Sink instead of gossipsub byte-identically; persist journal entry matches published raw; Close is a no-op on nil topic/subscription; sink error surfaces from Publish without recording pending.
• Existing cmd/lantern devnet-init tests updated: fake lotus stub now answers eth_chainId + Filecoin.Version so the boot-time discovery succeeds; new asserts check EthChainID = 31415926 + BlockDelaySecs = 4 land in the persisted config; discovery-call floor raised from 3 to 5.
• Full suite green: go test ./... reports 51 packages ok, no failures.

Not in this PR

• [Design Lantern brand mark: logo + wordmark #8/Peer count decays after boot; need 30s keepalive redial loop #9] ChainHead head-poll lag on devnet. Same root cause: single-node devnet can't form a gossipsub mesh so head updates come only from the fallback RPC poll (slow). Folding into a separate --devnet-head-poll-interval knob PR after this batch lands.
• [JWT tokens are issued without exp claims and never expire #7] eth_getCode proxy-address slowness on a specific PDPVerifier proxy address. The auto-wired VMBridge fix should already resolve this in practice (bridges through lotus, not empty local state), but keeping the finding open until a live smoke confirms.

Follow-up

Live devnet smoke deferred (colima needs restart + fresh docker compose up). Unit tests cover the parse + persist + handler-read paths end-to-end via an in-process fake lotus, but the truest test is against a running make docker/devnet node. I'll capture that in a follow-up comment on #123 once colima is back up.

[Reiers]

Live-smoke verification against running docker devnet

Booted a fresh docker compose up -d under projects/curio-fork/docker/ (lotus v1.35.1, BlockDelay=4, network localnet-1069c685-ccf0-49e4-8e2c-f2b4a4ed7f7e, chainId 31415926) plus lantern devnet-init + lantern daemon --network devnet on the same binary as this branch HEAD.

devnet-init discovery

✓ StateNetworkName: "localnet-1069c685-ccf0-49e4-8e2c-f2b4a4ed7f7e"
✓ Genesis:          bafy2bzaceahwbexzeywl2pd3hm2ggh5w7fhfibuf2v7v5wig5us363b75ku44
✓ ChainHead:        epoch 15538, state root bafy2bzacect3x7k7lxduvxnpkbrjj4ulki7rqfid7msnm64fwfd7ktq6m5m5y
✓ eth_chainId:      31415926 (0x1df5e76)
✓ BlockDelaySecs:   4

RPC parity (6 fixes verified live)

Method	Result	Was
Filecoin.StateNetworkName	localnet-1069c685-...-a4ed7f7e	"devnet"
Filecoin.Version.BlockDelay	4	30
Filecoin.ChainGetGenesis	genesis CID	not implemented
eth_chainId	0x1df5e76	0x13a
net_version	"31415926"	"314"
Filecoin.MpoolGetConfig	{PriorityAddrs: null, SizeLimitHigh: 30000, ...}	HTTP 500

Boot log includes:

  vm-bridge:    devnet auto-wire (lantern#123) → http://127.0.0.1:1234/rpc/v1
  vm-bridge:    forest@127.0.0.1:1234/rpc/v1

Partial (see finding #7)

eth_getCode on the real PDPVerifier proxy address (0x564Fce2BeCEdd3042F0Cb8bE787fe8230a40d276) reaches the auto-wired bridge but still times out at 15s. Lotus directly serves the same call in <1s. Root cause: eth_getCode tries local hamt walk (state cache → bitswap → gateway) before falling through to the bridge, and single-node devnet has no bitswap peers so the walk waits its full timeout. Fix candidate is a devnet-only short-circuit that goes bridge-first for these methods; out of scope for this PR. Filing a follow-up.

Follow-ups (out of scope, worth their own issue/PR)

• --devnet-head-poll-interval knob to close the head-lag findings (ChainHead + eth_blockNumber lag by ~80 epochs on single-node devnet since gossipsub can't propagate live head at 4s cadence)
• devnet-only bridge-first ordering for read-side eth_* methods (finding JWT tokens are issued without exp claims and never expire #7 in its narrower form)
• standalone lantern daemon CLI doesn't wire mpool at all today (this PR's mpool sink is in pkg/daemon for embedded consumers like curio-core). MpoolPushMessage on the standalone CLI has always been broken; adding baseline gossipsub mpool wiring to cmd/lantern (with devnet sink fallback) is a separate PR.