Improve qSPECT conversion and end-to-end validation reliability

[plesqui]

Summary

This PR strengthens the qSPECT DICOM conversion workflow and improves the reliability of the voxel dosimetry validation pipeline.

• Uses RealWorldValueMappingSequence as the sole quantitative mapping for qSPECT images, preventing double scaling from conventional rescale tags.
• Adds stricter validation and clearer errors for required DICOM metadata.
• Normalizes Siemens and GE manufacturer values.
• Adds an optional Siemens frame-duration fallback.
• Corrects signed 16-bit pixel metadata after qSPECT quantization.
• Improves injection activity decay correction and validates measurement chronology.
• Replaces the legacy SPECT-to-SUV tutorial with an updated qSPECT workflow.
• Updates the end-to-end validation test to use corrected RTSTRUCT assets from Zenodo.
• Validates Zenodo downloads using expected file sizes and MD5 checksums.
• Retries truncated RTSTRUCT and voxel-kernel downloads before failing.
• Adds tests for qSPECT scaling, DICOM metadata handling, radioactive decay, and interrupted kernel downloads.

Testing

• tests/test_dvk.py: 4 passed
• Formatting checks passed.
• git diff --check passed.
• Run qSPECT (DicomModify) on prior data and quantification (activity in Bq/mL) per voxel was identical.

[carluri]

Thank you @plesqui