Various fixes for XAI in 2.1

[GalyaZalesskaya]

Summary

Issue: CVS-142112

Found and fixed various issues:

• Dump saliency maps using label names, if possible
• The score_treshold for classification, detection and inst_seg is too low. This means that choosing saliency maps for only predicted classes works incorrectly. Added additional score_thr=0.3 to choose saliency map with high confidence
• Fix softmax for detection saliency map generation algorithm
• Fix saliency maps for YOLOX. Added cropping saliency maps if it was padded.
  YOLOX resizes the initial image, keeping the aspect ratio and padding the image to square. The generated saliency map is also padded so after overlaying it on the initial image, it looks distorted. Added cropping to fix it.
• Fix h-label cls saliency map dumping for predicted classes. Converted predicted labels from h-label format to index in labels_name list based on hierarchical label info, e.g. [0,0,1] -> 5 CVS-144040

TODO:

• Fix XAI for MobileNet V3
• Fix XAI for multilabel cls
• Apply these changes to Model API to update OV task

How to test

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have ran e2e tests and there is no issues.
• I have added the description of my changes into CHANGELOG in my target branch (e.g., CHANGELOG in develop).​
• I have updated the documentation in my target branch accordingly (e.g., documentation in develop).
• I have linked related issues.

License

• I submit my code changes under the same Apache License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below).

# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

[goodsong81]

Thank you for the fix. Please review my comments.