docs: use utilsforecast.evaluation.evaluate in notebooks

[IndarKarhana]

Closes #834

This PR replaces custom evaluate_performance functions and direct metric calls (e.g., mse(), mae()) with the standardized utilsforecast.evaluation.evaluate function across documentation notebooks.

Changes

• getting_Started_complete.ipynb

  • Added NIXTLA_ID_AS_COL=1 env var to ensure unique_id is a column (not index)
  • Replaced custom evaluate_cv implementation with evaluate(df, metrics=[metric], models=models)

• getting_Started_complete_polars.ipynb

  • Same evaluate() migration (polars already returns unique_id as column)

• GARCH_tutorial.ipynb

  • Added NIXTLA_ID_AS_COL=1
  • Replaced mae(cv_df, models=models) with evaluate(cv_df, metrics=[mae], models=models)

• CrossValidation.ipynb

  • Added NIXTLA_ID_AS_COL=1
  • Replaced rmse(cv_df, models=['AutoETS']) with evaluate(cv_df, metrics=[rmse], models=['AutoETS'], agg_fn='mean')

Why NIXTLA_ID_AS_COL?

The cross_validation() and forecast() methods currently return DataFrames with unique_id as the index. The evaluate() function expects unique_id as a column. Setting this env var ensures consistent behavior and prepares for the future default where unique_id will be a column.

Testing

• getting_Started_complete.ipynb passes nbdev_test
• Polars notebook unchanged (already compatible)
• Tutorial notebooks have .notest but changes follow same pattern

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codspeed-hq]

CodSpeed Performance Report

Merging this PR will not alter performance

_{Comparing IndarKarhana:issue/834-docs-evaluate-refactor (fe9b1e3) with main (0621049)}

Summary

✅ 6 untouched benchmarks

[nasaul]

LGTM