test: add end-to-end backend regression suite with real ASL fixtures and CI integration

[aviralsaxena16]

Hello @1brahimmohamed @jan-petr,

This PR introduces a structured end-to-end regression testing framework for the ASL Method Section Generator backend.

The objective is to improve reliability and establish a deterministic validation pipeline for API behavior and real BIDS-based ASL inputs, in alignment with the GSoC project goals.

All tests pass locally and are validated under CI across multiple Python versions. Snapshot baselines are version-controlled to ensure deterministic behavior across environments.

---

What This PR Adds

1. End-to-End API Regression Testing

Regression tests are implemented at the FastAPI layer using TestClient to validate:

• Root endpoint stability
• /api/report/process/bids behavior
• /api/report/process/dicom behavior
• /api/report/report-pdf generation

pytest-regressions is used to snapshot JSON responses, ensuring that future changes do not silently alter report generation logic.

Volatile fields (e.g., nifti_slice_number) are excluded to prevent false-positive regressions.

---

2. Real ASL Fixture-Based Testing

Added real ASL fixtures under:
tests/fixtures/real_sample/

Includes:

• sub-Sub1_asl.json
• sub-Sub1_m0scan.json
• sub-Sub1_aslcontext.tsv

Tests simulate real BIDS uploads including:

• JSON metadata
• ASL context TSV
• Valid NIfTI file

No internal functions are mocked , the full request lifecycle is exercised:

upload → normalization → metadata extraction → report generation → serialization

This validates:

• Upload pipeline integrity
• Metadata normalization logic
• Report generation consistency
• Error and warning propagation
• M0 consistency checks
• Slice extraction from NIfTI

This approach ensures scientific reproducibility by protecting report text and metadata transformations from unintended drift.

---

3. Deterministic Sample NIfTI

A minimal valid NIfTI file is included:

app/utils/sample_nifti.nii.gz

This allows CI to exercise nibabel-based slice detection without relying on external datasets while maintaining reproducibility.

---

4. API Schema & Error Handling Tests

Added coverage for:

• Missing file validation
• Invalid DICOM upload handling
• Correct HTTP status codes
• Structured error payload behavior

---

5. PDF Endpoint Validation

Verified:

• Successful PDF generation
• Correct application/pdf content-type
• Non-empty binary response

---

Minor Fixes Included

• Renamed modaliy_enum.py → modality_enum.py
• Updated imports accordingly

---

Impact

This PR establishes a reproducible testing baseline that enables:

• Safe refactoring
• Addition of new ASL variants
• Future modality expansion (DCE / DSC)
• Processor-level enhancements
• Metadata validation improvements

New fixtures can now be added per module with minimal setup.

Test Summary

(closes #7)