chore: bump CI Node versions to 22 and 24

[lodekeeper]

Summary

• Bump test matrix Node versions from [18, 20] to [22, 24] (Node 18 is EOL).
• Bump benchmark workflow Node version from 20 to 24 to match the release workflow.

Closes #495

Test plan

• CI green on test-node matrix for both Node 22 and Node 24
• Benchmark workflow runs on Node 24

🤖 Generated with AI assistance

[nflaig]

I believe this might need some repo setting changes to be able to merge it cc @wemeetagain

[lodekeeper]

Triaged before you get a chance to look at it — the blocker here isn't a repo setting. I checked:

• branches/master/protection → 404 (no classic branch protection)
• rulesets → [] (no rulesets)
• mergeable: true, mergeable_state: "blocked" is driven purely by the failing required status checks.

The actual blocker is two failing checks: Tests Node (22) and Tests Node (24). Both fail with the same symptom in packages/ssz spec-mainnet:

× mainnet/phase0/ssz_static/BeaconState/ssz_random case_0  16565ms
  → Test timed out in 15000ms.
× ... case_1 16023ms
× ... case_2, case_3, case_4 — all > 15000ms

Master's last Tests run (5ae98cf, Node 18/20) was green at 12:21 UTC, so it's specifically the big BeaconState/ssz_random cases crossing the 15s line on newer Node runners. Root cause looks like a latent bug in the test itself: test/spec/ssz_static.ts:53 calls vi.setConfig({testTimeout: 30 * 1000}) inside the it() callback, which doesn't take effect for the currently running test. Node 18/20 finished fast enough to stay under the default 15s; Node 22/24 pushed past it.

Fix is probably a one-liner — pass the timeout as the 3rd arg to it() or raise the default in vitest.base.unit.config.ts. Happy to push that here, or split it into its own PR first if you'd prefer to land the test fix independently before the Node bump. Let me know.

[nflaig]

@lodekeeper please fix the tests as part of this PR

[lodekeeper]

Pushed 9342349 — moves the mainnet timeout to it()'s 3rd arg in packages/ssz/test/spec/ssz_static.ts. The vi.setConfig({testTimeout: 30_000}) call was inside the it() callback and therefore didn't apply to the currently-running test, which is why the default 15s was still in effect and the BeaconState/ssz_random cases (16–17s on Node 22/24 runners) were timing out. Also dropped the now-unused vi import. CI is running — I'll watch for Tests Node (22)/(24) to go green.

[lodekeeper]

Heads up — the push landed action_required on Tests, Benchmark, and Labeler. The workflow runs need a maintainer to approve them before Node 22/24 will actually re-run against the fix. (This is probably the repo setting you were pointing at earlier — likely Settings → Actions → Require approval for all outside collaborators.) Could you (or @wemeetagain) kick off approval so the checks can run? I'll monitor once they start.