Investigate DRY approach for TypeScript/Python schema definitions

[vredchenko]

Context

From PR #113 discussion: we currently maintain schema definitions in two places:

• TypeScript types (core/*.ts) - for webui consumption, IDE type checking
• Python Pydantic models (packages/smartem-workspace/.../schema.py) - for CLI runtime validation

This creates maintenance overhead when schemas change.

Options to explore

1. JSON Schema as source of truth - generate both TS types and Python models from JSON Schema
2. Generate Python from TypeScript - tools like ts-to-python
3. Generate TypeScript from Python - datamodel-code-generator reverse mode
4. Keep separate - accept manual sync overhead

Acceptance criteria

• Evaluate each approach for complexity vs benefit
• Prototype preferred approach
• Document decision in ADR