test(e2e): ignore benign 'Insufficient valid txs' block-build-failed in epochs tests

[AztecBot]

Summary

Fixes the e2e_epochs/epochs_mbps "builds multiple blocks per slot with L2 to L1 messages" flake that hard-failed the merge-train/spartan CI in run 26150606646 (log 1779264300727828).

Root cause

The block builder emits block-build-failed with reason "Insufficient valid txs" whenever it attempts to build a block but the mempool happens to be empty at that exact moment (sequencer-client checkpoint_proposal_job.ts). In the failing run, sequencer-2 successfully built a 6-block checkpoint at slot 14, then tried to build a 7th block within the same slot after all 5 L2→L1 test txs had already been mined — the empty mempool produced the benign event.

EpochsTestContext.watchSequencerEvents already explicitly excludes block-tx-count-check-failed for exactly this reason (see existing comment on L512). block-build-failed with reason "Insufficient valid txs" is the same "not enough txs to build a block" case, just detected after the builder started processing rather than during the pre-check. The retry hit the same race, producing a hard CI failure.

Fix

In watchSequencerEvents, drop block-build-failed events whose reason is exactly "Insufficient valid txs". The other failure path in checkpoint_proposal_job.ts (the catch block) emits block-build-failed with the exception's message, so genuine block-builder errors still surface.

ClaudeBox log: https://claudebox.work/s/ed927253171f52af?run=1