Row filtering operation in pipeline

I wonder if there is an operator available to filter out rows from the training data based on tuneable parameters. I am only aware of the Selector-Transformer-Regressor steps, but I would like to build my own operator that for example:

- is able to remove (instead of replacing) outliers from the dataset with a tuneable threshold parameter
- or detects drifts in my time series training dataset and then discards old data

Is there currently a way or workaround to include such an operator in a pipeline?

In general I think the whole topic of drift detection, automatic retraining if needed is not addressed adequately in most AutoML frameworks, what you think?

No files, tests, or entry points are named. Start by reviewing how TPOT represents Selector-Transformer-Regressor pipelines and whether operators can remove rows or use tunable thresholds. Done would require a decided scope for row filtering or drift handling, an implementation path, and tests covering the supported behavior.

Written by the indexing model from the issue text.