feat: implement OCR regression harness with golden dataset and CI integration

[bytebinders]

✅ PR Description:

What was done:

• Modular Harness: Developed a full-featured OCR regression harness in app/ai-service/regression_harness/.
• Golden Dataset: Created a structured repository for "golden" documents and ground truth values in regression_harness/dataset/.
• Automated Evaluation: Implemented evaluator.py to compare actual OCR output against expected fields, supporting text normalization, error classification, and confidence tracking.
• CLI & Reporting: Added cli.py to run suites locally, providing human-readable console summaries and machine-readable JSON reports for CI artifacts.
• CI/CD Integration: Integrated a new GitHub Actions workflow .github/workflows/ocr-regression.yml to trigger automatically on OCR-related changes.
• Documentation: Provided a comprehensive README.md for tool usage, adding new samples, and maintenance.

Why it was done:
To establish a reliable, low-maintenance testing infrastructure that ensures OCR extraction accuracy is preserved as the AI models, prompts, or preprocessing steps evolve.

How it was verified:

• Verified the integrity of the data models and evaluation logic.
• Validated the CLI's reporting capabilities (JSON and Console).
• Verified the GitHub Actions configuration for environment dependencies (Tesseract-OCR).
• Local sanity check performed on the directory structure and file system operations.

Summary of Work:

1. Models: models.py defines the schema for samples and results.
2. Evaluator: evaluator.py implements the logic for field comparison and IoU calculation.
3. CLI: cli.py provides the interface to run tests and export results.
4. Dataset: Established ground_truth.json with sample documents.
5. CI/CD: Added ocr-regression.yml for automated regression testing.
6. Documentation: Created README.md.

Required line:
Closes #464

[drips-wave]

@bytebinders Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[bytebinders]

@Cedarich , could you drop a few more GitHub issues for me to work on? I’m ready for more 🔧🙂

[Cedarich]

Please fix work flow

[bytebinders]

@Cedarich the issue is missing file sampl_001.png but I have fixed it.

[Cedarich]

Fix workflow