fix(ci): fail health check when sitemap route checks fail

[dralgorhythm]

Summary

The "Check site availability" step in .github/workflows/health-check.yml could never fail on a broken sitemap route. The per-route loop was fed by a pipe:

echo "$CHECKS" | jq -c '.[]' | while read -r check; do ... FAILED=1 ... done

A pipeline runs the while in a subshell, so FAILED=1 set inside it never propagates to the parent shell — the final if [ $FAILED -eq 1 ] always saw 0. Only the robots.txt/sitemap.xml probes (outside the loop) and the Lighthouse gate were actually enforced, defeating the workflow's main purpose (including the scheduled uptime run).

Fix

Feed the loop via process substitution instead, keeping it in the parent shell:

while read -r check; do ... done < <(echo "$CHECKS" | jq -c '.[]')

Two-line change; ✅/❌ log output and all other behavior unchanged. Bash-only syntax is safe here — the step has no shell: key and GitHub's default run shell on ubuntu-latest is bash.

Verification

• actionlint (with embedded shellcheck): output byte-identical to the pre-change baseline — no new findings. (Shellcheck can't catch this bug itself: FAILED is also assigned outside the loop, which masks SC2031.)
• bash -n on the extracted step script: syntax OK.
• Ran the actual extracted step scripts (old from HEAD, new from this branch) under bash -e with a stubbed curl on PATH:
  • Old script, one route down → prints ❌ but exits 0 (bug reproduced)
  • New script, one route down → same log output, exits 1
  • New script, all healthy → all ✅, exits 0

🤖 Generated with Claude Code

[github-actions]

🔍 Preview Build Ready

Build Status: ✅ Success
Node Version: 24

📦 Preview artifact uploaded - check the Actions tab to download.

This preview was built from commit e3a34eb