[Pipeline] Updates

[zacandcheese]

Transformation

VDF Functions that affect and alter the entire table are currently not supported, only VColumn transformations are currently allowed.
These functions should be supported:

• case_when
• abs
• applymap
• swap
• rolling

Balancing the data and other preprocessing functions should be incorporated into pipelines.

• balance()
• .normalize()
• .fillna()
• .dropna()
• .fill_outliers()

Training

Non Native Functions should be supported. They should not raise an error and their underlying sql should be allowed to be omitted in the code.
These algorithms should be supported:

• Nearest Centroid
• MCA
• KNeighborsClassifier
• LocalOutlierFactor

In-Memory Models should also be investigated.

Only one model can be tested at a time, we may want to train multiple if computations become expensive.

Ingestion

Broaden the application and testing to include the following:

• created table
• verticapy datasets
• local load from file
• dataloader

Metrics

• Allow for classification metrics which are mainly python based with a sql call to make the contingency matrix to be run but not have it error for not adding a sql.

Tests

Transformation Tests

• Cyclic Tests
• Order of cols Tests

Training Tests

• include parameter
• exlclude parameter
• split parameter