Utilize new prediction pipeline

[aditya0by0]

Use new prediction pipeline as per ChEB-AI/python-chebai#135

[aditya0by0]

• The target labels for each trained model can also be included in lightning checkpoint. This avoid to manual bookkeeping of the labels file for each model-data variant. If this is agreed upon, we can create a separate issue for this.

• Right now how to handle the invalid smiles or None returns is unclear for ELECTRA due to ChemDataReader.read_data incorrectly accepts entire invalid strings with atleast one valid token python-chebai#137

[aditya0by0]

This is implement in ChEB-AI/python-chebai#135.

• The target labels for each trained model can also be included in lightning checkpoint. This avoid to manual bookkeeping of the labels file for each model-data variant. If this is agreed upon, we can create a separate issue for this.

TODO:

remove target_labels_path: classes.txt from registry
update readme
check if logic needs any change for classes enabled in checkpoing