Add CI-monitoring instruction to CLAUDE.md

[jnasbyupgrade]

Adds a ## GitHub CI section to CLAUDE.md directing Claude Code sessions to monitor GitHub CI after every push to a branch with an open PR (via gh pr checks --watch in a background subagent) and to fix failures immediately.

This mirrors the convention already documented in other Postgres-Extensions repos (e.g. cat_tools), so agent behavior is consistent across the org.

Docs-only change; no code or build impact.

[jnasbyupgrade]

@claude please review this PR.

[claude]

Claude finished @jnasbyupgrade's task in 46s —— View job

---

Reviewing PR #16

• Gather context (PR diff, CLAUDE.md)
• Review the added CI-monitoring section
• Post review feedback

Review

The diff is small and low-risk (docs only, +4/-0 in CLAUDE.md):

+## GitHub CI
+
+After pushing to a branch with an open PR, monitor CI using `gh pr checks --watch` in a background subagent until all jobs pass or a failure is confirmed. Investigate and fix failures immediately rather than leaving them for the user to notice.

Looks good:

• gh pr checks --watch is a valid gh CLI invocation that polls until all checks resolve — correct choice for this purpose.
• Scoping ("after pushing to a branch with an open PR") is precise and avoids triggering on unrelated pushes (e.g., pushing a branch with no PR yet).
• Directive matches the stated goal of aligning behavior with the cat_tools convention, and it's phrased consistently with the rest of CLAUDE.md's imperative style.

Minor nits (non-blocking):

• Placement: the new ## GitHub CI section is inserted at CLAUDE.md:5-8, ahead of ## Project Overview. Every other CLAUDE.md file in this repo family leads with a project overview before operational directives — worth considering moving this section lower (e.g., near "Development Workflow") purely for reading order, though it has no functional effect since Claude Code loads the whole file regardless of order.
• "fix failures immediately" is unqualified — for a repo with test_factory__owner role/security-definer bootstrapping (per this file's own Architecture section), some CI failures could stem from environment/setup issues rather than code bugs; might be worth a caveat that "fix" means "fix or clearly report if the failure is environmental," but this is a very minor wording point and not a correctness problem.

No functional/build impact — this is purely instructional text for future Claude Code sessions. I'd consider this ready to merge as-is; the nits above are optional polish.