feat(sequencer): add waitForBuildDeadlineOnFinalBlock flag

[PhilWindle]

What

Adds an opt-in sequencer config flag, waitForBuildDeadlineOnFinalBlock (default false), that makes the final block of a checkpoint keep collecting txs until its build deadline instead of sealing as soon as minTxsPerBlock are available. Enables it in e2e_epochs/epochs_mbps_redistribution to make that test deterministic.

Why

epochs_mbps_redistribution dumps a burst of late txs and asserts they all land in the checkpoint's final block (new Set(lateBlockNumbers).size === 1). The proposer releases the final block as soon as getPendingTxCount() >= 1 and then iterateEligiblePendingTxs takes a single mempool snapshot. The burst is submitted with Promise.all(...send()) and reaches the proposer's pool over gossip one tx at a time, so the snapshot captures only a subset and the rest spill into the next checkpoint — an intermittent failure, not a redistribution bug.

Rather than change production block-sealing (which would idle the final block or eat into execution time), this gates the wait behind a flag that only tests enable:

• Default (false): unchanged — every block, including the final one, seals as soon as it reaches minTxsPerBlock.
• true: the final block of a checkpoint waits until its build deadline (block_build_deadline − min_block_duration) before sealing, so a late burst is fully present at the snapshot. The build deadline already reserves the post-build window for attestation collection, so checkpoint timing is unaffected.

Changes

• stdlib/src/interfaces/configs.ts: new optional SequencerConfig field + schema entry.
• sequencer-client/src/config.ts: default (false) and CLI mapping (no env var — testing-only, like skipInvalidateBlockAsProposer, so it does not appear in the operator CLI reference).
• sequencer-client/src/sequencer/checkpoint_proposal_job.ts: thread isLastBlock into waitForMinTxs; when the flag is set, the final block collects until its build deadline.
• end-to-end/src/e2e_epochs/epochs_mbps_redistribution.test.ts: enable the flag.

Testing

• New checkpoint_proposal_job.timing.test.ts cases for both behaviours (seal-early by default; wait-to-deadline when set).
• sequencer-client unit suites pass (timing 23/23, job 46/46) and the package typechecks. The e2e itself runs in CI (no docker locally).

Follow-up

Supersedes the flake marking in #24098 for this test. Once both land, the epochs_mbps_redistribution entry added to .test_patterns.yml by #24098 can be removed.

[AztecBot]

Flakey Tests

🤖 says: This CI run detected 1 tests that failed, but were tolerated due to a .test_patterns.yml entry.

\033FLAKED\033 (8;;http://ci.aztec-labs.com/49af6f91d30d1de9�49af6f91d30d1de98;;�): yarn-project/end-to-end/scripts/run_test.sh ha src/composed/ha/e2e_ha_full.parallel.test.ts "should distribute work across multiple HA nodes" (228s) (code: 0)