fix: tolerate unavailable redis in ci3 logging

[AztecBot]

Summary

Fixes the nightly debug build wrapper failure from https://github.com/AztecProtocol/aztec-claude/actions/runs/28845951015.

Root cause: redis_setexz piped input through gzip | redis_cli -x ... even when CI_REDIS_AVAILABLE=0. In that mode redis_cli returned without reading stdin, so gzip hit SIGPIPE and the strict set -euo pipefail path aborted before bootstrap_ec2 could launch the build.

Changes:

• Make unavailable Redis a real no-op that drains stdin for redis_setexz.
• Make unavailable redis_cli -x drain stdin as well, covering the direct cat ... | redis_cli -x SETEX ... path.
• Add ci3/tests/redis_test for the no-Redis pipefail behavior.

Verification

• ci3/tests/redis_test
• env CI=1 CI_REDIS_AVAILABLE=0 bash -c 'set -euo pipefail; NO_CD=1 source ci3/source; echo "CI booting..." | redis_setexz test-key 1'
• bash -n ci3/source_redis ci3/tests/redis_test
• git diff --check

./ci3/bootstrap.sh also ran the new test successfully, then failed in existing ci3/tests/signal_test because Docker is unavailable in this container and that test's EXIT cleanup trap returns nonzero after its skip path.

---

Created by claudebox · group: slackbot

[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/08e48c7b19de7492�08e48c7b19de74928;;�): yarn-project/kv-store/scripts/run_test.sh src/sqlite-opfs/internal/ordered-binary-browser.test.ts (2s) (code: 0)