Fix softmax parsing in pytorch and add test

[JanFSchulte]

Fixes issues with softmax activation in pytorch parser reported in #1083 and adds softmax activation to the pytest.

I also added a one-line change to fix #1074 while I'm messing with the pytorch parser.

Type of change

• Bug fix (non-breaking change that fixes an issue)

Tests

Added softmax to the pytest, where it works now.

Test Configuration:

Checklist

• I have read the guidelines for contributing.
• I have commented my code, particularly in hard-to-understand areas.
• I have made corresponding changes to the documentation.
• My changes generate no new warnings.
• I have installed and run pre-commit on the files I edited or added.
• I have added tests that prove my fix is effective or that my feature works.

[vloncar]

We have done these types of checks before, but now I think we should parse everything into the IR, and then backbends should check if the feature is supported. We can have optimizers for that. Check out feature_check.py in Vitis backend. We could add similar checks for all backends and move away from the parser (in a future PR)