fix(archiver): tolerate re-included already-stored checkpoints

[spalladino]

Context

Reconciling checkpoints from L1 could permanently wedge the archiver. BlockStore.addCheckpoints accepts an already-stored checkpoint that an L1 reorg re-included in a different L1 block (same archive root, new L1 metadata), but it filtered the already-stored prefix only locally: ArchiverDataStoreUpdater.addCheckpoints still re-ran log and contract-data extraction for every block in the original batch. Re-presented blocks carrying a ContractClassPublished or ContractInstancePublished event then threw already exists, aborting the whole write transaction — the L1 sync point never advanced, so every retry re-fetched the same range and failed deterministically until manual DB cleanup.

Approach

• BlockStore.addCheckpoints now returns the checkpoints it actually inserted (excluding the already-stored prefix handled by skipOrUpdateAlreadyStoredCheckpoints) instead of a boolean, and the data store updater derives the blocks for log/contract-data extraction from that suffix.
• As defense in depth, duplicate contract class/instance inserts are now a no-op when the stored insertion block number matches the incoming one; a duplicate at a different block number still throws, since that signals genuine double-processing.
• Regression tests cover re-including an already-stored checkpoint alone and batched with a new one, plus store-level tests for the same-block/different-block duplicate behavior.

Reported in AztecProtocol/aztec-claude#1288.

Fixes A-1350

[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/39e8d89bef20799e�39e8d89bef20799e8;;�):  yarn-project/end-to-end/scripts/run_test.sh simple src/multi-node/slashing/duplicate_attestation.test.ts (73s) (code: 0) group:e2e-p2p-epoch-flakes