Add rigid debug timing and release DCCCcore 4.2.4

[tctco]

Summary

• add an RAII DebugReporter for structured stage timings
• instrument rigid loading, preprocessing, ONNX inference, iterative convergence, debug image output, and final save
• document large CTA .nii.gz compression as a separate benchmark cost
• release DCCCcore 4.2.4 without the previous -alpha suffix
• point dcccpy and the Linux ARM64 vendor helper at the 4.2.4 GitHub release

Behavior

The registration algorithm and default non-debug path are unchanged. Timing output is enabled only when the rigid command receives --debug. The ONNX output tensor setup also replaces a temporary shape-vector pointer with a named local shape whose lifetime covers tensor creation.

Validation

• clean-first Docker build: all 60 DCCCcore C++ objects compiled and linked
• rebuilt binary reports 4.2.4
• maximum referenced glibc version: 2.16
• real 8 MB PET NIfTI completed rigid --iterative --debug, converged after two iterative passes, and produced a valid output NIfTI
• dcccpy tests: 11 passed, 1 skipped
• UI metric command tests: 4 passed
• git diff --check passed