fix(examples): compute the discrete-label threshold from the training split only

[Amir Fathi (AmirF194)]

Why are these changes needed?

In the "Forecasting Discrete Values" example, above_mean_sales is thresholded on df["Sales"].mean() computed over the full 180-day series, before the train/test split happens a few lines later. That mean is derived in part from the 30-day period held out as discrete_test_df, so the training labels carry information from the data the model is later evaluated against.

This PR splits into discrete_train_df / discrete_test_df first, computes the threshold from discrete_train_df["Sales"] only, and applies that single value to both frames. Same fix applied to both the notebook and its mirrored markdown doc under website/docs/Examples/.

Verification:

• Reproduced the leak with the notebook's own data generator (hcrystalball.utils.get_sales_data, same call the cell uses): the full-series mean (17020.14) differs from the train-only mean (16833.40) and changes at least one training row's label.
• Ran the fixed cell logic standalone (same library calls, outside the notebook) and confirmed discrete_X_train/discrete_X_test keep their original shapes (150/30 rows) and discrete_train_df keeps its original column set.
• Did not re-run the notebook's downstream AutoML.fit/predict cells (no stored outputs exist for the cell I changed, and no CI test executes this notebook), so I can't speak to any effect on the reported accuracy number further down, only that the label computation itself no longer touches the test period.
• Checked the docs mirror (website/docs/Examples/AutoML-Time series forecast.md) for the same pattern and fixed it identically; ran pre-commit on both changed files (mdformat reflowed two lines in the doc, otherwise clean).

Related issue number

Fixes #1175

Checks

• I've used pre-commit to lint the changes in this PR.
• I've included the doc change needed (the markdown mirror of this notebook cell had the same leak).
• I've added tests (no test in this repo executes automl_time_series_forecast.ipynb; the fix was verified by running the cell's logic standalone, see above).
• I've made sure all auto checks have passed.

[Amir Fathi (AmirF194)]

@microsoft-github-policy-service agree