Use fork-aware sync message due point in beacon node callers

[shane-moore]

Description

#9675 added SYNC_MESSAGE_DUE_BPS_GLOAS and a fork-aware ChainSpec::get_sync_message_due_at_slot, but only the VC consumes it. Three BN call sites still use the fork-blind get_sync_message_due():

• lighthouse/beacon_node/beacon_chain/src/light_client_optimistic_update_verification.rs

  Line 74 in 1cd1a3c

  let sync_message_due = chain.spec.get_sync_message_due();

  and

  lighthouse/beacon_node/beacon_chain/src/light_client_finality_update_verification.rs

  Line 78 in 1cd1a3c

  let sync_message_due = chain.spec.get_sync_message_due();

  - the TooEarly gossip gates. The spec defines these via get_sync_message_due_ms(), which Gloas modifies. Post-Gloas we would reject light client updates from spec-compliant peers arriving between 25% and 33.33% of the slot, and downscore them.

• lighthouse/beacon_node/beacon_chain/src/validator_monitor.rs

  Line 1539 in 1cd1a3c

  spec.get_sync_message_due(),

  - grades sync message delay against the pre-Gloas due point. Metrics only.

Harmless today since GLOAS_FORK_EPOCH is far-future in every config.

Steps to resolve

Switch the three call sites to get_sync_message_due_at_slot, using the update's signature_slot for the light client gates and the message's slot in the validator monitor. I think signature_slot is the right fork selector since the gate measures time from the start of that slot, but a test at the fork boundary would pin it down.

Once wired, get_sync_message_due() has no production callers and can be removed, renaming get_sync_message_due_at_slot to get_sync_message_due::<E>(slot) to match get_attestation_due.

Additional Info

Part of the broader inner-slot timing work in #8615.

[NikhilSharmaWe]

@shane-moore I can take this. Could you assign it to me?

[NikhilSharmaWe]

This looks done in #9823.

The LC TooEarly gates and the validator monitor now use get_sync_message_due::<E>(slot) with signature_slot / the message slot, and the fork-blind getter is gone.