Add unit tests for coverage_curve()

[jc-macdonald]

Summary

Unit tests for coverage_curve() — empirical coverage across nominal levels.

Closes #6

Changes

• Return types: returns two numpy arrays
• Default levels: 50 evenly spaced points from 0.05 to 0.99
• Custom levels: accepts user-specified level array
• Monotonicity: empirical coverage is non-decreasing
• Well-calibrated: posteriors drawn from same distribution give coverage near the diagonal (max deviation < 0.15)
• Bounded: all empirical values in [0, 1]
• Degenerate posteriors: point mass at truth gives 100% coverage at all levels

Stats

• 7 new tests in tests/test_scoring.py (50 total)
• scoring.py coverage: 0% → 44%
• Total coverage: 67.30% → 68.34%
• just ci passes (format, lint, mypy strict, coverage)