refactor: Use compiled json for schema in production
* Use dynamic import for schema based on environment
* If production use pre-generated json to avoid loading vega generator -- big memory footprint reduction
* If not production use vega so schema is always up to date