Fix the error report for a demand forecast

An AI evaluated a demand forecasting model with R-squared alone, reported 0.91 and picked the model with the highest score. In production the forecasts were consistently off in one direction. R-squared rewards following the ups and downs, and it is blind to a steady bias. The fix reports MAE, WAPE and bias instead.

An assistant wrote forecast_error_report(actuals, preds). It is in the editor. It has a bug. Find it and fix it.

Input and output:

• actuals: the units actually sold each day. preds: the forecast for each day, at the same index.
• Return a list of three numbers: MAE, WAPE and bias.

Rules the tests check:

• Use only the first n days, where n is the shorter length. If n is 0, return an empty list.
• A day's error is its forecast minus its actual.
• MAE: the mean of the absolute errors.
• WAPE: the sum of the absolute errors, divided by the sum of the absolute actuals. If that sum is 0, WAPE is -1.
• Bias: the mean of the errors, with their signs. A positive bias means the model forecasts too high.
• Answers within 0.000001 of the expected values pass.