Move HampelFilter from transformations to annotation module

Is your feature request related to a problem? Please describe. The HampelFilter was included in the transformation module prior to the addition of the annotation module. It is used to detect outliers, which is an annotation task.

I’ve talked with @aiwalter who added it, and we both think it should be moved to the annotation module in the long-run. But we don’t want to move it without being able to still use it (and other annotators) in a ForecastingPipeline.

Describe the solution you’d like HampelFilter should be moved to the annotation module and refactored to comply with that interface. This should be done in conjunction with completion of #840, which will add a ConditionalImputer transformer that will allow annotators to be used in a ForecastingPipeline.

Describe alternatives you’ve considered The alternative of leaving the HampelFilter in the transformation module is not ideal in the long-run because then we have estimators for the same task (detecting outliers) in more than one module.

Additional context N/A

[satya-pattnaik]

but what do you need to do according to exog?

Please pardon me, I confused Z=DataFrame with exog. You can ignore that.

[aiwalter]

Yes I’ll push some changes by coming Monday, and I might need help in adding the exogenous part as mentioned by @aiwalter and to write up some unit tests.

Sure lemme know, but what do you need to do according to exog?