test(e2e): fix expiration_timestamp prune race + flake-group e2e_slashing

[AztecBot]

Summary

Re-enables the merge queue on merge-train/spartan after PR #23344 was dequeued by three test flakes in the ci/x5-full heavy-grind shard (run 26169283519). None are caused by #23344's diff.

What failed

Test	Error	Log
src/e2e_expiration_timestamp.test.ts	Block header not found in invalidates-the-transaction setup (line 147)	http://ci.aztec-labs.com/97be94f0ae50ea78
src/e2e_p2p/sentinel_status_slash.parallel.test.ts	Expected checkpoint-invalid, got checkpoint-missed	http://ci.aztec-labs.com/9d83b09229998a31
src/e2e_slashing/attested_invalid_proposal.test.ts	Timeout awaiting honest validator slash offenses	http://ci.aztec-labs.com/7f9ad9eedbf86aee

Fix — e2e_expiration_timestamp (real fix, not just patterning)

Root-caused to a real pipelining race. The with no enqueued public calls / invalidates the transaction test calls cheatCodes.eth.warp(provedExpiration + aztecSlotDuration, { resetBlockInterval: true }), jumping L1 ~37 slots ahead. The proposer had pre-built block 9 for slot 10 (pipelining), but L1 is now past slot 10 without that block being checkpointed, so the archiver prunes it (Pruning blocks after block 8 due to slot 10 not being checkpointed). The prune is asynchronous and lands during the next test cycle. That next test's beforeEach reads getBlockData('latest') cleanly, then 3ms later runInvalidatesTest reads it again and races the in-flight prune — resolveBlockTag('proposed') returns block 9's number, but stores.blocks.getBlockData({ number: 9 }) returns undefined because the row was just deleted.

Fix: cache the header timestamp from beforeEach and reuse it in runInvalidatesTest instead of refetching mid-cycle. The first read tolerates the warp's reorg because it's after a fresh test transition; the second read was redundant.

Flake patterns — e2e_slashing

The attested_invalid_proposal timeout (5min budget) under heavy grind isn't a missing-hash race; it's slot-counting timing. Adds src/e2e_slashing/.*\.test\.ts to the existing e2e-p2p-epoch-flakes group (owner: palla), alongside the existing p2p/epoch patterns. The e2e_p2p/sentinel_status_slash failure is already in that group via the src/e2e_p2p/.* regex.

Group threshold stays at 5, so up to 4 simultaneous flakes are tolerated.

Details in https://gist.github.com/AztecBot/3600f093693517fa69c8981f6f31dbe7

[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/dd176d3d03d9ee15�dd176d3d03d9ee158;;�):  yarn-project/end-to-end/scripts/run_test.sh simple src/e2e_p2p/sentinel_status_slash.parallel.test.ts "slashes the proposer with INACTIVITY when checkpoint validation records unvalidated" (205s) (code: 0) group:e2e-p2p-epoch-flakes