Experiment further with Kuzu DDL for prompt engineering

[prrao87]

The test suite in tests/test_text2cypher_ddl_schema.py need more digging in and experimentation. We could potentially look for better ways to directly use Kuzu's DDL to prompt the Text2Cypher models. However, the challenge is structured outputs of high enough quality when generating free strings via the DDL prompt in BAML - unlike the JSON output (where BAML handles the prompt formatting for us), we'd need more prompt engineering to get this to work as well as the JSON schema approach.

Starter script to debug and run queries are in query_graph_ddl_schema.py.