Update Bass notebook to use BassModel class

[williambdean]

Update docs/source/notebooks/bass/bass_example.ipynb to showcase the new BassModel class.

Motivation

The current notebook uses create_bass_model directly. Since the refactor (#2584) added the BassModel(ModelBuilder) class as the recommended entry point, the notebook should demonstrate the class-based workflow.

Sections to update

1. Import — Use from pymc_marketing.bass import BassModel as primary entry point
2. Model creation — BassModel(model_config=...) instead of separate priors dict
3. Fitting — model.fit(data=...) with data in various formats (array, DataFrame, xr.Dataset)
4. Posterior analysis — Use model.idata directly with ArviZ
5. Posterior predictive — model.sample_posterior_predictive(X=future) for forecasting
6. Save/Load — model.save() / BassModel.load() round-trip
7. Plotting — model.plot_* methods (once # is complete)
8. MLflow — pymc_marketing.mlflow.autolog(log_bass=True) (once MLflow integration for BassModel #2588 is complete)

Notes

• Keep the existing simulation + multi-product structure — it's a good demo
• Replace pm.sample calls with model.fit where possible
• The create_bass_model function should still be mentioned as a lower-level alternative

[williambdean]

Connected issues:

• Plotting methods for BassModel #2589 — Plotting methods (notebook should use these)
• MLflow integration for BassModel #2588 — MLflow integration (notebook should showcase)
• Dual y-axes for the "Innovators vs Imitators" plots in bass notebook #1706 — Dual y-axes for innovator vs. imitator plots