fix(ci): make aztec-up verdaccio prime retry patiently

[AztecBot]

Why

The merge-train/spartan-v5 train PR (#23975) was dequeued from the merge queue. The merge-queue-heavy CI failed in grind run x8-full (ci log) at the aztec-up task (log):

npm error 404 Not Found - GET http://localhost:4873/@aws-sdk%2fcredential-provider-env - no such package available
npm error 404  The requested resource '@aws-sdk/credential-provider-env@^3.972.46' could not be found
Failed after 3 attempts: npm i -g --prefix /tmp/npm-prime @aztec/aztec@0.0.1 @aztec/cli-wallet@0.0.1 @aztec/bb.js@0.0.1

The aztec-up test primes a local verdaccio registry by running npm i -g @aztec/aztec ..., which resolves every transitive dependency fresh from npmjs (proxied through verdaccio). The 404'd version @aws-sdk/credential-provider-env@3.972.46 was actually published ~13h before the run, so this was a transient verdaccio↔upstream-npm fetch miss, not a missing package — and unrelated to the three commits in the train (#23967, #23821, #23976). The aws-sdk dependency tree resolves cleanly again now.

The default retry helper (3 attempts, 5s apart) only covered ~2 min, which wasn't enough to ride out the blip, so the prime failed and the train was dequeued.

What

Retry the verdaccio prime up to 8 times with 30s backoff before failing. The happy path is unchanged (first attempt breaks immediately); only failures pay the extra patience.

This is CI resilience hardening, not a code fix in the train. The same change should be ported to next (the non-v5 aztec-up/bootstrap.sh is identical).

---

Created by claudebox · group: slackbot

[AztecBot]

Automatically closing this stale claudebox draft PR (no updates for 5+ days). Re-open if still needed.