[TASK] Make pairwise losses support 3D predictions tensors and Keras Masking

[gabrielspmoreira]

Description

The pairwise losses were originally tested with 2D prediction tensors (batch size x # items/classes). For sequential/session-based recommendation, the prediction tensors are 3D (batch size x seq length x # items/classes), which are currently not supported. In addition, the pairwise losses are not considering the Keras masking of prediction tensors (introduced in #775 ), so that loss is only computed for masked positions of the sequential targer.

• [ ] Pairwise losses - Add support to prediction tensors with rank greater than 2 (e.g. 3D)
• [ ] Pairwise losses - Add support to Keras Masking of predictions tensors