feat: add deterministic quality scoring engine with tests

[yuliuyi717-ux]

/claim #1

Implemented a deterministic multi-dimensional quality scorer for structured submissions with required output schema, benchmark coverage, and sample scorecards.

What is included:

• quality_scorer.py
  • auto-detects json, markdown, code, text
  • scores 5 dimensions with required weights:
    • completeness 0.30
    • format_compliance 0.20
    • coverage 0.25
    • clarity 0.15
    • validity 0.10
  • returns required schema:
    • weighted_score
    • quality_rating
    • scores
    • feedback
    • pass_threshold
• tests/test_quality_scorer.py
  • format detection tests
  • output schema test
  • weighted-score consistency test
  • 100 submissions performance test (<10s)
  • ground-truth tolerance test (mae <= 0.05)
  • sample scorecards file integrity test
• sample_scorecards.json
  • 20 sample scored outputs
• scripts/generate_sample_scorecards.py
  • regenerates the sample scorecards

Validation:

• python3 -m unittest discover -s tests -p 'test_*.py'

[yuliuyi717-ux]

Quick verification note:

• python3 -m unittest discover -s tests -p 'test_*.py' passes locally.
• The suite includes format detection, schema validation, weighted-score consistency, and a 100-submission benchmark check under 10s.
• Included sample_scorecards.json with 20 generated scorecards plus a generator script for reproducibility.

If you want stricter calibration checks against your provided 20-item ground-truth set, I can wire that in directly.

[yuliuyi717-ux]

Follow-up update pushed:

• added ground-truth calibration utility (evaluate_ground_truth_submission_set) to compute MAE + tolerance pass/fail
• added CLI script (scripts/evaluate_ground_truth.py) to validate against a provided 20-item truth set
• added test coverage for tolerance gating

Validation:

• PYTHONPATH=. python3 -m unittest discover -s tests -p 'test_*.py'

This directly strengthens the acceptance item around +/-0.05 calibration checks.