fix(ci3): tolerate unavailable redis in log writes

[AztecBot]

Summary

• Make redis_setexz drain stdin and return success when Redis is unavailable.
• Suppress noisy docker ps output from the local dashboard probe when Docker is not reachable.
• Add a focused regression test for the unavailable-Redis pipe path.

Root Cause

The linked barretenberg-claude Actions run is not readable from this public-mode session, but the same nightly debug launcher failure is reproducible from the current CI helper: with CI_REDIS_AVAILABLE=0, echo "CI booting..." | redis_setexz ... exits under pipefail with SIGPIPE before bootstrap_ec2 can launch the debug build.

That means the nightly can report as a barretenberg debug build failure even though no C++ build or test has started. The helper now consumes stdin when Redis is intentionally unavailable, so callers that use redis_setexz for best-effort log publication do not abort.

Verification

• CI=1 CI_REDIS_AVAILABLE=0 bash -lc 'set -euo pipefail; source ci3/source; echo "CI booting..." | redis_setexz test-key 1; echo after'
• ci3/tests/source_redis_test
• bash -n ci3/source_redis ci3/tests/source_redis_test .github/ci3.sh ci.sh ci3/bootstrap_ec2

---

Created by claudebox · group: slackbot