[Feature]: Add comprehensive test suite for NaFlex vision embedding processor

🚀 Feature Description

Add comprehensive test coverage for the NaFlex vision embedding processor in kornia/models/processors/naflex.py, which currently has 0 test coverage despite being a core utility function.

📂 Feature Category

VLM/VLA Models (Vision Language Models/Agents) - Priority

💡 Motivation

• NaFlex processor has 0 tests (274 lines of implementation, 0 test lines)
• Core utility for vision embedding processing with flexible resolution has no verification
• No tests for image preprocessing functionality
• No gradient checks or torch.compile compatibility tests
• Users have no reference for proper processor usage
• No validation of output format for vision models

Why this is needed:

• Verify image preprocessing pipeline works correctly
• Ensure flexible resolution handling produces correct outputs
• Test processor compatibility with PyTorch features (autograd, torch.compile)
• Prevent regressions in vision embedding processing
• Provide usage examples for integrating with vision models

💭 Proposed Solution

Add comprehensive test suite covering NaFlex vision embedding processor:

Test Coverage Plan

Core functionality tests:

• ✅ Smoke tests: Basic instantiation with different configurations
• ✅ Exception tests: Invalid inputs, edge cases (invalid resolutions, wrong tensor formats)
• ✅ Cardinality tests: Output shape verification for various input resolutions
• ✅ Flexible resolution tests: Verify handling of arbitrary input sizes
• ✅ Image preprocessing: Test normalization, resizing, padding operations
• ✅ Batch consistency: Verify batch processing produces same results as individual processing
• ✅ Gradient checks: Verify backpropagation correctness using gradcheck
• ✅ Torch.compile compatibility: Test with torch_optimizer fixture
• ✅ Integration tests: Test with actual vision model pipelines

Target metrics:

• 70-90 test cases covering all scenarios
• ~230+ lines of test code (similar to implementation size)
• All tests following BaseTester pattern

Additional Documentation

Jupyter notebook demonstrating:

1. Processor instantiation and configuration
2. Image preprocessing pipeline walkthrough
3. Flexible resolution handling examples
4. Integration with vision-language models
5. Performance benchmarks for different input sizes
6. Visual demonstration of preprocessing steps

🔄 Alternatives Considered

No response

🎯 Use Cases

For developers:

• Verify processor works correctly after code changes
• Test compatibility with new PyTorch versions
• Understand preprocessing pipeline internals

For users:

• Learn how to use NaFlex processor with vision models
• Understand flexible resolution capabilities
• See practical integration examples
• Get copy-paste examples for custom pipelines

For maintainers:

• Ensure preprocessing quality
• Catch regressions in vision embedding processing
• Reduce maintenance burden from integration issues

📝 Additional Context

• Location: kornia/models/processors/naflex.py
• Use case: Vision embedding processor with flexible resolution
• Purpose: Preprocessing utility for vision models

🤝 Contribution Intent

• I plan to submit a PR to implement this feature
• I'm requesting this feature but not planning to implement it