Move files into the organization required

[Kaftow]

This PR refactors the subject migration pipeline by extracting the shared classification logic out of scripts/migrate_subjects.py into reusable modules.

What changed

• Added core/json_loader.py to centralize loading of mapping JSON files from scripts/mappings/
• Added rule_engine/normalization.py for shared normalization helpers and regex-based detection of reading levels and classification codes
• Added core/subject_classifier.py as the reusable core for subject classification at both subject and work level
• Simplified scripts/migrate_subjects.py so it acts primarily as the CLI / orchestration layer instead of owning all classification logic
• Updated scripts/README.md to document the new internal module structure

Why

Previously, the migration script contained mapping loading, normalization utilities, and classification logic all in one file. This made the code harder to reuse, test, and extend. By moving those responsibilities into dedicated modules, the branch makes the classifier easier to maintain and prepares the codebase for reuse outside the CLI script.

Impact

This is primarily a structural refactor. The intended classification behavior remains the same, but the implementation is now split into clearer layers:

• data loading
• normalization / low-level detection
• reusable classification logic
• CLI entry point

[Kaftow]

The CLI remains the entry point, but the reusable tagging logic now lives in a package-style layout:

• tagging/__init__.py for the public re-export surface
• tagging/engine.py for the TypedTagger implementation
• tagging/normalization.py for pure normalization and classification helpers
• tagging/json_loader.py for mapping JSON I/O
• tagging/models.py for typed result structures