fix: require structured challenge test results

[saurabhhhcodes]

📌 Description

• Type: Security / Bug Fix / Tests
• Summary: Challenge submissions now derive correctness from structured judge results instead of trusting stdout text containing Passed. JavaScript submissions with problem test cases get an internal per-case judge marker, all expected cases must report passed: true, and failed case details are stored with the submission.

🔗 Related Issues

• Closes Challenge correctness is derived from fragile output text #194

✅ Checklist

• Code compiles and runs clean
• Added/Updated documentation
• Added/Updated tests
• Linted and formatted code
• Related issue linked
• No sensitive data added

📸 Screenshots (if applicable)

Not applicable; backend challenge evaluation fix.

💬 Notes for Reviewers

The regression test covers the exploit case directly: stdout can contain Passed, but the submission remains incorrect unless the structured judge markers cover every configured test case and the execution exits cleanly.

🧪 How to Test This PR

cd backend
npm install --ignore-scripts
node --check controllers/challenge.controller.js
node --check models/ChallengeRoom.js
node --check tests/challenge.test.js
npm test -- challenge.test.js
npm test
cd ..
git diff --check

Local result: challenge suite passed with 8 tests; full backend suite passed with 13 files and 75 tests.

📦 Tech Stack

• Node.js
• Express
• MongoDB
• Vitest

[vercel]

@saurabhhhcodes is attempting to deploy a commit to the Sameer's projects Team on Vercel.

A member of the Team first needs to authorize it.