migrate core processing to utilsforecast

[jmoralez]

• Bumps the min python version to 3.8
• Moves polars to be an optional dependency
• Adds deprecation warnings to passing df to the StatsForecast constructor.
• Makes use of functions from utilsforecast to perform the processing.
• Uses the base grouped array from utilsforecast to avoid some code duplication.
• Deprecates the sort_df argument (in utilsforecast we sort if needed).
• Only saves the original dates and times if we'll compute the fitted values.
• Removes the warnings about ds being object and tries to parse it as it previously did.
• Removes the usage of df in the constructor in the fugue methods.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[AzulGarza]

This is great, @jmoralez! Abstracting GroupedArray in utilsforecast is a really cool effort. The code now is more navigable. 🎉

I just had a question regarding using unique_id as an index.

[AzulGarza]

We agreed on maintaining the warnings for more time to allow users to change their code. 🙌