Add multi-frame PET motion correction

[tctco]

Summary

• add a general PetMotionCorrector preprocessing module for X × Y × Z × N PET NIfTI data
• add DCCCcore pet-motion-correct with optional corrected-dynamic NIfTI and motion TSV outputs
• independently register every frame to frame 0 using ITK Euler 3D rigid transforms, Mattes mutual information, multi-resolution registration, and linear resampling
• automatically motion-correct and average declared 4D inputs before the existing adni-pet-core 3D pipeline, while leaving declared 3D inputs on the existing path
• distinguish true 3D NIfTI from declared 4D N=1 data, despite ITK collapsing trailing singleton axes
• update DCCCcore to 4.3.0 and document the command

Tests

• Docker Release build: passed
• full Docker pytest suite: 113 passed, 8 skipped
• targeted post-review rebuild/tests: 5 passed
• manually verified /home/cheng/DCCCSlicer/localizer/src/tests/4frames.nii.gz: 4 frames corrected and averaged successfully

New automated coverage includes:

• 3D input rejection by pet-motion-correct
• 4D output-to-3D conversion and float output type
• declared 4D N=1 identity behavior
• known synthetic translations/rotations with substantial alignment recovery
• corrected dynamic output and motion TSV contents
• unchanged declared-3D adni-pet-core routing
• automatic 4D motion correction → average → ADNI PET Core integration