Description: Changed port for batch extraction to avoid loop in case of LSTM. Now MO takes batch from the first Parameter. In original Kaldi there is no batch, we add it in MO and have not change anywhere during graph transformations, so using batch from Parameter is valid approach for Kaldi and allows to reshape stateful networks by batch.

Ticket: 23956

Code:

• Comments
• Code style (PEP8)
• Transformation generates reshape-able IR
• Transformation preserves original framework node names Not applicable, no new transformations

Validation:

• Unit tests
• Framework operation tests Not applicable, no new operations
• Transformation tests Not applicable, no new transformations
• e2e model test with an update of ./tests/e2e_oss/utils/reshape_utils.py Should be done in separate PR
• Model Optimizer IR Reader check Not applicable, no new operations

Documentation:

• Supported frameworks operations list Not applicable, no new operations
• Supported public models list Not applicable, no new models
• New operations specification Not applicable, no new operations
• Guide on how to convert the public model Not applicable, no new models
• User guide update Not applicable, have not mentioned that Kaldi models can't be with batch > 1