Replace the hand-built metadata subset in parseMeta with a VR-aware
DICOM-to-DICOM-JSON converter. Window Center/Width, Rescale Intercept/Slope
are only emitted when the source element exists; multi-values are preserved
in source order. VOI/LUT, Presentation LUT and Pixel Padding tags are
preserved and Pixel Data is excluded.
- add src/dimse/dicomJson.ts (dictionary-aware converter)
- rewrite src/dimse/parseMeta.ts to use the converter
- add synthetic non-PHI DICOM writer and metadata tests
- add npm test script (node:test + ts-node) and skipLibCheck
