Fix dtype precision, improve depth scaling tolerance, add debug logging, update manifest weights, enhance preprocessing output.

Add validation utilities, update model conversion logic, fix manifest.json, rename deprecated flags, improve docs

Fix core ml precision mismatches, update quaternion handling, add global scale storage, enhance detailed model validation, improve multi-image support, fix metadata UUIDs, enable sRGB color output conversion.

Add quaternion validator with configurable tolerances, enhance coreml validation with detailed stats, add PIL-based image preprocessing, introduce batch/image-specific validation functions, update CLI flags for input/output handling.

git lfs stuff