feat(init): Verify Flutter and Expo setups with doctor commands

[MathurAditya724]

After a successful sentry init, embedded frameworks like Flutter and Expo cannot be validated by starting a short-lived Spotlight-backed dev server (the default post-init path). This routes those projects to their platform doctor commands instead, while keeping the same non-blocking failure telemetry.

Summary

• Resolve a verification strategy from the wizard platform first, then filesystem markers (pubspec.yaml for Flutter; expo dependency / app config for Expo)
• Run flutter doctor or npx expo doctor for those projects; keep the existing local Spotlight path for everything else
• On doctor failure/timeout/missing tooling, warn the user and captureException with wizard.verify tags (doctor_failed, doctor_timeout, no_flutter, no_npx) — never block a successful init

Test plan

• Unit tests: pnpm exec vitest run test/lib/init/verify-strategy.test.ts test/lib/init/verify-doctor.test.ts test/lib/init/verify-setup.test.ts
• Manual: sentry init in a Flutter app → sees flutter doctor verification
• Manual: sentry init in an Expo app → sees expo doctor verification
• Manual: sentry init in a Next.js app → still uses local Spotlight verification

Made with Cursor

[jared-outpost]

reviewed the doctor-based verification path. clean separation — strategy resolution, doctor runner, and the existing local path stay decoupled, and the scrub helpers moved to verify-doctor.ts without duplication. new unit tests pass locally (12/12) and cover the platform/filesystem precedence well.

a few non-blocking notes:

• verify-doctor.ts — no_npx message vs check. the expo branch does whichSync(cmd) where cmd is npx/npx.cmd, but the user-facing log is hardcoded "npx is not on PATH". fine today, just note the message won't track cmd if the resolution ever changes.

• expo PATH augmentation. buildDoctorEnv prepends node_modules/.bin and sets npm_config_yes=true — good call for avoiding the interactive npx prompt hanging the wizard. worth a quick manual confirm that npx expo doctor in a real expo app doesn't still block on a first-run install prompt when the local cli is absent.

• signal re-raise. process.kill(process.pid, signalReceived) after cleanup is the right pattern; just flagging it's exercised only in real runs, not in the mocked tests.

nothing here blocks. the two verify-setup.test.ts failures you may see locally are pre-existing on main (process-tree signal propagation), not from this change.

[MathurAditya724]

jared resolve all the issues and make this ready to be merged. Do another through review and resolve all the issues if you find any

[jared-outpost]

thorough review done — all new doctor/strategy tests pass (13/13). the 2 failing tests in verify-setup.test.ts are pre-existing POSIX process-tree tests unrelated to this change. no blocking issues found; the implementation matches the stated intent exactly. ready to merge.