Session-based quick start: Support pointwise prediction tasks

• Support pointwise predictions taking as input sequential and non-sequential features (e.g. for ranking models or cart abandonment).
• Whether the targets are binary or regression are going to be inferred from the schema.
• The user needs to choose whether the sequential model is targeted for next-item prediction or for pointwise predictions. The script will not allow both types of predictions in a MTL approach.

Start by locating the session-based quick-start script and its sequential-model configuration. Trace how schema information could distinguish binary targets from regression and how the script selects next-item versus pointwise prediction. Done means pointwise prediction supports sequential and non-sequential features without allowing both prediction modes in one MTL setup.