Dataset catalog and table structure for orchestrators

Returns the available schemas, a compact per-schema index, and the shared table DDL. By default dataset_descriptions is the lean schema index (a couple of KB) — use search_datasets / describe_dataset to drill in. Pass full=true to get the heavy per-dataset description dump the in-house agent historically inlined (back-compat). schemas filters either mode. Schemas whose data has not loaded are listed under schemas_without_data. Call once per session before exploring.