feat: E2E test suite for api mode

[ArthurCRodrigues]

Context

The Autograder required a comprehensive E2E test suite to ensure the stability of the entire submission lifecycle across all grading templates in a live environment. Previous tests were largely unit-based or used mocks, which didn't capture infrastructure-related issues or integration bugs between the API, Sandbox Manager, and the various templates.

Solution

This PR introduces a robust E2E test suite that validates the Autograder in a live Docker environment:

1. E2E Test Suite: Implemented a new testing directory tests/e2e with 20+ scenarios covering:
   • Submission Lifecycle: Creation, polling, detail retrieval, and history verification.
   • Template Coverage: Exhaustive testing for input_output, web_dev, and static_analysis (5+ scenarios each).
   • Assignment Management: Full CRUD operations for grading configurations.
   • M2M Ingestion: Verification of external result ingestion (e.g., from GitHub Actions).
2. Environment Orchestration: Added scripts/run_e2e_tests.sh to automate building and starting the Docker environment, running tests, and cleaning up.
3. Infrastructure Fixes:
   • Lazy Docker Initialization: Refactored SandboxManager to lazy-load the Docker client, preventing API crashes in remote mode when Docker is not locally available.
   • Integration Auth: Configured default integration token in docker-compose.yml to support authorized requests during E2E runs.
4. Diagnostics: Improved test output with detailed state printing on failures to facilitate debugging in CI environments.

Further clarifications

• api_testing template scenarios are currently marked as xfail due to known networking constraints in the sandbox environment that require separate architectural changes.
• AI-based tests are skipped unless OPENAI_API_KEY is provided in the environment.

Related issues

Closes #302

Checklist

• I linked the related issue(s) and explained the motivation.
• I kept this PR focused and scoped to a single concern.
• I added or updated tests for changed behavior.
• I ran the relevant tests locally.
• I updated documentation when needed.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[ArthurCRodrigues]

ok