feat(snapshot): build shadowfork snapshots from a jochemnet release

[qu0b]

Adds a workflow that turns a jochemnet benchmark release into six snapshots for a shadowfork. They include the release's deployed contracts and start at the same block, with a lower gas limit.

1. Verify the release bundle, replay it on Geth, and lower the gas limit to 200M by default.
2. Replay the same blocks on the other five clients.
3. Compact Geth and Erigon through their own commands. Explicitly skip compaction for Nethermind, Besu, Reth and Ethrex; no external database tools or manual file swaps.
4. Export the datadirs and update latest after all six exports succeed. Recover each runner's scratch data afterward.

Geth must reopen at the expected head. Nethermind uses snapshot-specific persistence settings and must pass an offline reopen check before export, including its head hash and state root.

After the benchmarkoor action cleans up, each job restores the CLI from the same cached image for compaction and export.

Dependencies (merged): benchmarkoor #349, #350 and #351.

Validation: workflow YAML, shell syntax and ShellCheck pass. CLI extraction was checked with a Docker stub for supplied/local images and cleanup on success/failure. Earlier end-to-end builds and the finalized mock used the previous compaction policy; the revised workflow has not had a full six-client run.

[redpandabot]

Summary

The PR replaces the old shadowfork-image-build workflow/scripts with benchmarkoor-snapshot.yaml, four helper scripts and a jochemnet snapshot builder config that replay a release, ramp the gas limit down, compact where a client command exists, and publish per-client snapshots. The latest commit correctly drops the external ldb compaction (nethermind/besu/reth/ethrex now skipped, geth/erigon use their pinned client image), and the earlier out-of-scope shadowfork-file and SNAPSHOT_PREFIX findings are gone. One blocker from the previous reviews remains: the benchmarkoor binary the workflow calls for compaction and export is deleted by the benchmarkoor action before those steps run.

Issues

• 🔴 .github/workflows/benchmarkoor-snapshot.yaml:203 — benchmarkoor CLI is deleted by the action before db compact / snapshot export — see the thread on that line

---

_{Reviewed @ 28ac19d4
"First law of programming: whatever the bug, blame DNS."}

[redpandabot]

🔴 benchmarkoor CLI is deleted by the action before db compact / snapshot export

The ethpandaops/benchmarkoor@master steps (lines 146 and 263) are a composite action whose final Cleanup step runs rm -f "${{ runner.temp }}/benchmarkoor" (benchmarkoor action.yaml:455, which is the same $RUNNER_TEMP/benchmarkoor the action's Extract benchmarkoor binary step wrote at line 135). By the time the Compact steps (line 206 and line 291) and snapshot-export.sh (line 22) run, that binary no longer exists, so every client fails with No such file or directory before any archive is published. Re-extract the binary (or run the CLI from the image) after the action step, before compacting/exporting.

[redpandabot]

The PR adds benchmarkoor-snapshot.yaml plus five helper scripts and a builder config that turn a jochemnet release into six shadowfork snapshots, and the follow-up commit under review restores the benchmarkoor CLI from the already pulled/built local image after the composite action deletes it. All previously reported findings are resolved: geth/erigon compaction now passes the pinned --image "$SNAPSHOT_IMAGE", the unused --compaction-image/SNAPSHOT_PREFIX paths are gone, the shadowfork/ files those earlier findings referenced were removed by this branch, and both jobs re-extract the binary before compact/export. I found no remaining real problem in the change.

---

_{Reviewed 8 changed file(s) @ d7da397d — no blocking issues found.
"First law of programming: whatever the bug, blame DNS."}