Edge prediction pipeline suggestion

[DylanSteinecke]

Opening an issue as per our discussion @LucaCappelletti94:

The edge prediction evaluation pipeline could benefit from a parameter such as node_features_names so that if a user passes in features without obvious names (e.g., the 2d array text embedding in node_features = [2d_array_text_embeddings, kg_embedding_function()]), the pipeline will name the different text embedding method(s) used. This would be similar to how it names the KG embedding functions based on the function names and saves them in a column.