[RMP] Support of advanced sequential tasks and the definition of examples

[sararb]

Problem:

Following the integration of session-based support (#451, #452), the missing component would be providing recsys practitioners with all the necessary blocks for advanced use-cases such as multi-task learning for item-level predictions and negative sampling support for the NextItemPrediction task.

Goal:

• Implement the blocks and negative sampling strategies required for training with item-level prediction tasks.

Constraints:

• [Task] Support of session-based model definition with masking techniques for TF #452 should be finalized before starting this work

Starting Point:

• Refactory of the prediction blocks to support the ContrastiveLearning class.

• Extend regression and classification prediction block to support sequential tasks (item-level)

• Implement negative sampling support for session-based tasks
  Note: support bpr, bpr-max, top1, and top1-max as defined in the GRU4Rec paper.

• Definition of advanced example notebooks showcasing multi-task learning for the session-based recommendation.