fix: address valid AI quality findings

[rhuanbarreto]

Summary

Reviewed all 12 AI findings from GitHub's Code Quality AI scanner and triaged them by validity. This PR addresses the ones worth fixing:

• Fix PT-BR typo: Corrected truncated word coment → comentários in the Sentry study translation table
• Improve plugin install observability: Added logDebug() calls to the three silent catch blocks in tryInstallPlugin() (Cursor, Copilot, Claude) so auto-install failures are visible in debug mode while preserving the intentional fall-through to manual URLs
• Add detail to missing credentials: The { installed: false } return now includes a detail explaining why installation was skipped

Dismissed findings (not addressed)

Finding	Reason
Replace Bun.env with process.env in paths.ts	Violates ADR ARCH-014 — project mandates Bun.env in all src/ files
Add max depth to findProjectRoot	False positive — dirname() terminates at filesystem root (parent === dir), no infinite loop possible
4x test coverage suggestions for url.test.ts	Wrong scope — structural unit tests don't need integration-level URL output verification
Use randomUUID in test uniqueId	Cosmetic — Date.now() + Math.random() is adequate for test isolation

Test plan

• bun run validate passes (lint, typecheck, format, 603 tests, ADR check, build)
• ARCH-014 ADR check specifically passes (no process.env in src/)

🤖 Generated with Claude Code

[cloudflare-workers-and-pages]

Deploying archgate-cli with    Cloudflare Pages

Latest commit:	b0face8
Status:	✅  Deploy successful!
Preview URL:	https://f2e2cd4c.archgate-cli.pages.dev
Branch Preview URL:	https://fix-ai-quality-findings.archgate-cli.pages.dev

View logs