feat(settings): generate repo onboarding pack for accepted Gittensory repos

[JSONbored]

Parent phase: #236
Parent roadmap: #127
Cross-link: repo policy compiler #277

Problem

Accepted repos should be able to publish a concise onboarding pack that tells miners what lanes are useful, what is discouraged, how maintainers want PRs shaped, and where Gittensory context fits.

The onboarding pack should consume normalized policy output from #277 instead of duplicating policy parsing logic.

Acceptance criteria

• Onboarding pack includes repo policy, contribution lanes, validation commands, public-output boundaries, and maintainer expectations.
• Generated text is previewable before export or publication.
• Public output passes sanitizer rules.
• Policy/compiler output from feat(settings): add repo policy compiler for focus manifests #277 can feed onboarding-pack generation.

Validation expected

• Tests cover generation and sanitizer boundaries.
• Fixtures prove compiler output can become onboarding-pack input.
• Screenshots show preview/export flow if UI is touched.

UI evidence gate

Any visible web, browser-extension, or GitHub-overlay change must include maintainer-reviewable screenshots or a short recording covering the changed states. A checked template box without actual visual evidence is not enough.

Public-output safety criteria

• Public text is sanitized before reaching GitHub comments, issue bodies, PR bodies, extension-visible public panels, or copied public snippets.
• Tests cover forbidden wallet/hotkey, reward-estimate, trust-score, public-score-prediction, private-reviewability, private-scoreability, and farming-language leakage.

Cross-cutting acceptance criteria

• Preserve the repo quality gate: npm run test:ci, 97%+ global coverage, and the local branch coverage target for touched code.
• Keep public/private boundaries explicit. Public GitHub output must not expose wallets, hotkeys, reward estimates, raw trust scores, public score predictions, private reviewability, private scoreability context, or farming language.
• Add/update focused tests for the changed behavior instead of relying on green checks alone.