make_databricks_loader(spark, catalog) in runner.py prefixes schema.table
refs with the catalog so narwhals expressions resolve to 3-level
catalog.schema.table on Databricks without parameter naming changes.
88 integration tests validate every pipeline transpiles to valid
Databricks SQL via sqlglot, checking parse validity, catalog prefixes,
output modes, and absence of DuckDB temp view artifacts.
3abb77904e
