Corrections in MO to fix performance drop in Interpolate-4 when mode=linear instead of linear_onnx.
Ticket: #49045

Code:

• Comments
• Code style (PEP8)
• Transformation generates reshape-able IR: Yes
• Transformation preserves node names: Yes

Validation:

• Unit tests
• Framework layer tests: N/A because there are no new supported layers
• Transformation tests: N/A
• e2e model test: N/A because there is no new model
• MO IR Reader tests: N/A because there is no new model

Documentation:

• Supported frameworks layers list: N/A because there are no new operations
• Supported public models list: N/A because there is no new public model
• New operations specification: N/A because there are no new operations
• Guide on how to convert the public model: N/A because there is no new public model
• User guide update: N/A because there is no need to update common user guide

Other:

• Sample/Demo application to infer the model: N/A because there is now new model