ci: consolidate AI model downloads and add retry logic

[Vladush]

Description

This PR addresses the frequent CI failures caused by unstable connections when downloading the YuNet and SFace .onnx models from GitHub and HuggingFace.
Instead of each build job downloading the models independently (5 parallel downloads), a new download-models job fetches the models once, and passes them to the build jobs via workflow artifacts.

In addition, scripts/download_models.sh has been upgraded to include a robust retry mechanism with exponential backoff (up to 3 attempts) and atomic file writes, which prevents partial/corrupted files from passing size checks.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update

How Has This Been Tested?

• Verified the retry logic and exponential backoff trigger correctly when simulating a network failure.
• Verified that models are successfully skipped when already present.

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• 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 added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes