Auto-detect feature layer for Pytorch models

[goodsong81]

[Changes]

• Feature layer auto detection (by lazy one-time forward hook)
  • CNNs: the last layer which outputs 4D maps w/ spatial dimension
  • ViTs: the N-last LayerNorm layer
    (Different from OV logic as there is no explicit Add module just before the LayerNorm in Pytorch models)

[Future works]

• Refactoring + torch optional install
• Documentation + example
• Functional tests

[codecov]

Codecov Report

Attention: Patch coverage is 96.07843% with 2 lines in your changes missing coverage. Please review.

Project coverage is 94.23%. Comparing base (8b0ddf9) to head (6617ccb).
Report is 1 commits behind head on develop.

Files with missing lines	Patch %	Lines
openvino_xai/methods/white_box/torch.py	96.07%	2 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop      #64      +/-   ##
===========================================
+ Coverage    94.16%   94.23%   +0.07%     
===========================================
  Files           27       27              
  Lines         1817     1857      +40     
===========================================
+ Hits          1711     1750      +39     
- Misses         106      107       +1     

Flag	Coverage Δ
dev-py310	94.23% <96.07%> (+0.07%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[negvet]

Validation on a large scope of models is the best way to see how such kind of changes generalize. I see you plan to do it as the future work.