Add forecasting article to pkgdown site

Add at least one practical pkgdown article showing how to use `bcchr` data in a forecasting workflow without expanding the package core API.

Suggested scope:

- Retrieve one monthly Banco Central series with `get_series()`.
- Convert the returned tibble to a regular time-series object only inside the article.
- Fit a simple forecast model such as `forecast::auto.arima()`.
- Produce a short forecast with uncertainty intervals.
- Plot observed and forecast values.
- Keep `forecast` in `Suggests`, not `Imports`.
- Do not add `forecast.bcch_series()` or other S3 methods; the article should demonstrate composition with existing packages instead.

The goal is to make the pkgdown site useful beyond API reference while keeping `bcchr` focused on discovering, describing, and retrieving Banco Central data.

Inspect the existing pkgdown article structure and package metadata first, then use get_series() to retrieve a monthly Banco Central series and compose it with forecast::auto.arima(). Done means the site contains a practical article with conversion, forecasts, uncertainty intervals, and a plot, while forecast remains in Suggests and no new S3 methods or core API are added.

Written by the indexing model from the issue text.