Implement the cl-pretrainer training

[afmjoaa]

• [x] Implement construct_batches_for_cl_pre_trainer inside the batch builder
• [x] Implement the training loop and return the batch accuracy and batch loss on each epoch.
• [x] Add checkpoint manager for cl pre trainer model
• [x] Add test for training and saving the model

[afmjoaa]

Only train for next word prediction task.

[afmjoaa]

Done