Fix seed boolean null handling

[nkwork9999]

Summary

Fixes #5589.

This preserves NULL values in seed boolean columns instead of coercing them to False.

The issue was caused by seed boolean conversion calling str_to_bool(str(value)) for every value. Pandas reads CSV null values as NaN, so the seed renderer converted NaN to the string "nan", and str_to_bool("nan") returned False.

Changes

• Preserve pandas NA values as None when rendering seed boolean columns.
• Add a regression test for a seed CSV containing null, false, true, and null in a boolean column.

Databricks validation

Verified on Databricks Free Edition / SQL Warehouse that Databricks preserves boolean NULL values with the expected SQL shape:

CREATE SCHEMA IF NOT EXISTS sqlmesh_smoke;

CREATE OR REPLACE TABLE sqlmesh_smoke.seed_bool_null_smoke (
  id STRING,
  test_ind BOOLEAN
);

INSERT INTO sqlmesh_smoke.seed_bool_null_smoke
SELECT
  CAST(id AS STRING) AS id,
  CAST(test_ind AS BOOLEAN) AS test_ind
FROM VALUES
  ('1', NULL),
  ('2', FALSE),
  ('3', TRUE),
  ('4', NULL) AS t(id, test_ind);

SELECT
  id,
  test_ind,
  test_ind IS NULL AS is_null
FROM sqlmesh_smoke.seed_bool_null_smoke
ORDER BY id;

Result:

1  null   true
2  false  false
3  true   false
4  null   true

Also verified locally that SQLMesh now generates Databricks SQL preserving NULL values:

FROM VALUES
  ('1', CAST(NULL AS BOOLEAN)),
  ('2', FALSE),
  ('3', TRUE),
  ('4', NULL) AS t(id, test_ind)

Tests

uv run pytest tests/dbt/test_transformation.py::test_seed_boolean_nulls_are_preserved -q
1 passed

uv run pytest tests/dbt/test_transformation.py::test_seed_boolean_nulls_are_preserved tests/dbt/test_transformation.py::test_seed_single_whitespace_is_na tests/dbt/test_transformation.py::test_seed_partial_column_inference -q
3 passed

uv run ruff check sqlmesh/core/model/definition.py tests/dbt/test_transformation.py
All checks passed!