Implement point-forecast metrics: RMSE, MAE, coverage

[jc-macdonald]

Summary

Utility metrics for point forecasts and prediction interval calibration.

Functions

Function	Description
rmse(obs, pred)	Root mean squared error
mae(obs, pred)	Mean absolute error
coverage(obs, lower, upper)	Empirical coverage of prediction intervals
coverage_curve(obs, quantile_levels, quantile_values)	Coverage as a function of nominal level

Design Notes

• These are not proper scoring rules but are commonly used alongside them.
• Group under a Diagnostics or Metrics submodule/section.
• coverage_curve is useful for calibration assessment (PIT histograms, reliability diagrams).

Acceptance Criteria

• RMSE/MAE match known values
• Perfect intervals give coverage = 1.0
• coverage_curve returns (nominal_levels, empirical_coverages) pairs