test: add guard training pipeline coverage

[joicyroslin-svg]

Closes #239

Summary

Adds focused test coverage for the Guard safety prediction training pipeline utilities.

Changes Made

• Added tests for training dataset preprocessing.
• Verified missing/invalid dataset columns raise a clear error.
• Added config loading test to ensure YAML overrides merge correctly with default training settings.
• Covered label normalization, text cleanup, duplicate removal, and reproducibility-related config behavior.

Testing

• pytest backend/tests/test_guard_training_pipeline.py -v

Result: 3 passed.

[joicyroslin-svg]

Hi @SdSarthak, I reviewed all failing CI checks for this PR.

This PR only adds one backend test file:

backend/tests/test_guard_training_pipeline.py

I verified the added test locally:

pytest backend/tests/test_guard_training_pipeline.py -v

Result: 3 passed.

Current CI failures appear unrelated to this PR:

• Backend Tests are failing in existing RAG/Auth tests such as test_rag_ingest.py, test_rag_query.py, test_rag_stream.py, test_rag_feedback.py, and test_rag_pipeline.py.
• Pytest auto-discover is also failing due to existing RAG integration tests returning 404 for RAG endpoints.
• Frontend Lint & Build is failing in existing frontend files like Login.tsx, Register.tsx, and api.ts because of unrelated lint/type issues.

Please let me know if you’d like any change in this PR, but the current failures seem outside the scope of the added guard training pipeline test coverage.