📝 IQR outlier detection at the start of a timeseries

[enryH]

No description provided.

[sambra95]

I tested a set of different outlier detection methods (see outlier_detection.ipynb) on some real growth curves that had some outliers and I also added several synthetic outliers (Test data and labels are stored in test_labels.csv and test_traces.csv).

I refactored the outlier detection logic to have a single detect_outlier function that then takes a method argument to decide which method to use. I put you iqr based method (method=iqr) and two other promising methods (ecod and hampel) as options. ECOD means another dependency but on the other hand it is both much more accurate and >50x faster than original method (see my Teams message)

I have updated the tutorial accordingly.