feat(history-expiry): save only headers for pre-merge blocks

[Gabriel-Trintinalia]

Description

This pull request introduces a new feature to enable saving only pre-merge block headers during synchronisation, along with related enhancements to the synchronisation pipeline and blockchain storage.

New Feature: Save Pre-Merge Headers Only for Snap Sync

• Added a new CLI option --Xsnapsync-synchronizer-pre-merge-headers-only-enabled to enable saving only pre-merge block headers during synchronization
• Introduced the SavePreMergeHeadersStep class to handle the logic for storing pre-merge headers in the blockchain.
  • Pre-merge block headers are stored and excluded from subsequent processing stages.
  • Post-merge block headers are passed through to the next processing stage.

Blockchain Storage

• Added a new method unsafeStoreHeader in MutableBlockchain to store block headers updating the chain state.

Synchronisation Pipeline Updates

• Updated FastSyncDownloadPipelineFactory to include SavePreMergeHeadersStep in the pipeline when the new feature is enabled.

Configuration Update

• Updated Sepolia checkpoint to the merge block

Pipeline before the changes:

(headers: blue | bodies: red)

Pipeline with this PR:

(headers: blue | bodies: red)

• Headers will be downloaded and saved until checkpoint block.
• After checkpoint, bodies and receipts will be saved as usual

Disk usage

Node Name	Checkpoint Block Number	Blob Blockchain Space
8582-headers-ctl-1	n/a	404 GiB
8582-headers-ctl-2	n/a	404 GiB
8582-headers-ctl-3	n/a	404 GiB
8582-headers-ctl-4	n/a	404 GiB
8582-headers-only-1	1,735,371	403 GiB
8582-headers-only-2	1,735,371	403 GiB
8582-headers-only-3	1,735,371	403 GiB
8582-headers-only-4	1,735,371	403 GiB
8582-sync-near-head-1	8,000,000	15 GiB
8582-sync-near-head-2	8,000,000	15 GiB

Syncing logs:

2025-04-27 09:13:05.264+10:00 | EthScheduler-Services-40 (saveHistoricalHeaders) | INFO  | SavePreMergeHeadersStep | Pre-merge block headers import progress: 100 of 50000 (0.20%)
2025-04-27 09:13:35.420+10:00 | EthScheduler-Services-40 (saveHistoricalHeaders) | INFO  | SavePreMergeHeadersStep | Pre-merge block headers import progress: 21500 of 50000 (43.00%)
2025-04-27 09:14:03.865+10:00 | EthScheduler-Services-40 (saveHistoricalHeaders) | INFO  | SavePreMergeHeadersStep | Pre-merge block headers import completed at block 50000 (0x0e30a7c0c1cee426011e274abc746c1ad3c48757433eb0139755658482498aa9)
2025-04-27 09:14:05.711+10:00 | EthScheduler-Services-43 (importBlock) | INFO  | ImportBlocksStep | Block import progress: 50111 of 22352512 (0%), Peer count: 2
2025-04-27 09:14:35.951+10:00 | EthScheduler-Services-43 (importBlock) | INFO  | ImportBlocksStep | Block import progress: 76800 of 22352512 (0%), Peer count: 2

[pinges]

One minor comment :-)

[siladu]

Thanks for the testing and write up 👍

[siladu]

Potentially a breaking change if any private networks make use of the existing checkpoint config?

We're kind of overloading the checkpoint concept, which I think makes sense for public networks as long as we're explicit about it. We should check usage for private.
In full 4444s, with rolling window, I guess this static checkpoint goes away entirely for public nets?

[siladu]

Missing changelog too - unless that's intentional?

[siladu]

LGTM. Just the changelog missing potentially, especially if we want to mention breaking changes (e.g. to private networks) at this point.