Tabular classification can only accept DataFrame objects

[franchuterivera]

When providing a numpy array object, we face the problem:

• The sklearn transformations like column transformer expect a 2D array
• When yielding an element from the base dataset, which is not pandas, we yield a 1D row from a tabular classification causing crashes.

The solution is to create a transformation for the tabular data loader that expand dimensions if the input is 1D via np.expandim.