Enable conversions to LSTM/GRU/RNN Sequence in MOC (#13470)
* Enable SequenceFusion, TensorIteratorToSequence, GRUCellFusion transformation in MOC, the transformations have been deleted from plugin pipelines
* clean-up: delete augru transformation header
* Temporary disabled TensorIterator IR reader tests
* temporary disable tests, it's not TI issue, usqueeze doesn't have reference impl for fp64 and bf16
* Update ConvertTensorItearatorToSequence transformations: add Unsqueeze to pattern
* Codestyle
* Specify negative values for LSTM/GRU/RNN Sequence ops
* fix conflicts with master branch
* Insert ShapeOf subgraph in ConvertSequencesToTensorIterator transformation in case of dynamic input shape
* codestyle
* fix conflict with master branch
* fix onednn version
* Update low latency v2 to support GRU/RNN/LSTM Sequence ops, fix accuracy issue
* fix tensor names
* fix tensor name issue on gna
* fix transformation tests, fix copying rt_info in the transformation
* fix warning, enable bf16, f64 ti tests
* codestyle
* fix functional tests
* EliminateDuplicateTIInputs transformation, fix copy_with_new_inputs method of TI op
* move EliminateDuplicateTIInputs transformation
* codestyle
5c0225d358