fix(studio): stop onboarding detect-site refetch loop on failed detection

[bnz183]

What

The onboarding wizard's detect-site step refetched /api/setup/detect in a tight loop whenever detection failed or returned nothing — ~1,600 requests in 3s, verified at runtime. React's dev-only passive-effect loop guard is stripped in production, so prod silently hammers the API rather than crashing.

Root cause

The effect keyed its guard on report === null, which conflated "not fetched yet" with "fetched and failed". On a null result report stayed null while setLoadingDetection(false) flipped a dependency, re-running the effect → guard passed → refetch, forever.

Fix (OnboardingWizard.tsx only)

• Track attempted separately from result via a resolvedAttempt token, set only when a fetch settles (success or failure). The guard resolvedAttempt === retryCount then stops re-fetching — null no longer doubles as "unattempted".
• Effect deps are [step, retryCount, resolvedAttempt] — no self-referential report/error outputs, so the effect can't re-arm itself.
• loading is derived (avoids react-hooks/set-state-in-effect); cleanup ignore flag drops stale/duplicate responses.
• Real error state with a Try again control; Continue is disabled on loading/error, closing the old "click through a blank wizard" hole.

Tests

New e2e/onboarding-detection-loop.spec.ts:

• Loop guard: force detect → 500, assert ≤1 request while idle, error notice visible, Continue disabled. (Was RED on the old code: "saw 377".)
• Retry: exactly one additional request and recovery on success.

Existing happy-path detection test (smoke.spec.ts) still green. tsc -b + targeted lint clean; no new lint suppressions.

Scope

Touches only OnboardingWizard.tsx and the new spec. The five other review findings on this branch are intentionally left untouched.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a376feb673

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Block continuing after a null post-save refresh

If the initial scan succeeds but the follow-up fetchSetupDetection() in handleApplySettings returns null after saving settings, detectionError is never set and resolvedAttempt still matches retryCount, so the wizard is in a loaded/null state. This disabled check treats report === null as okay, leaving Continue enabled and allowing users to advance into blank confirmation steps with no active suggestion; please either preserve the previous report or treat this null refresh as an error/blocked state.

Useful? React with 👍 / 👎.