fix: alias filler column in expression_is_true for strict-alias dialects

[sdebruyn]

Problem

default__test_expression_is_true emits an unnamed literal 1 as its placeholder column when should_store_failures() is false:

https://github.com/dbt-labs/dbt-utils/blob/main/macros/generic_tests/expression_is_true.sql#L7

{% set column_list = '*' if should_store_failures() else "1" %}

This compiles fine standalone, but as soon as anything wraps the rendered SQL in a subquery, strict-alias dialects reject it because the inner column has no name. Concrete failure modes:

• T-SQL (Microsoft SQL Server / Fabric Data Warehouse): Msg 8155, Level 16, State 2: No column name was specified for column 1 of 'sub'.
• Snowflake (historical): SQL compilation error: Missing column specification — reported in Missing column specification when using expression_is_true test (snowflake) #822, closed as Stale without a fix.
• Some Spark configurations that enforce strict aliasing in nested SELECTs.

Subquery wrapping is not exotic — it happens whenever a downstream macro composes generic tests, when stored-failures alternates with non-stored-failures, or when users manually wrap the test (for example to feed it into a custom monitoring model).

Fix

One-line change: name the placeholder column.

-{% set column_list = '*' if should_store_failures() else "1" %}
+{% set column_list = '*' if should_store_failures() else "1 as filler_column" %}

The name filler_column matches the convention already used in at_least_one, which carries an explicit comment about this same T-SQL constraint:

https://github.com/dbt-labs/dbt-utils/blob/main/macros/generic_tests/at_least_one.sql#L36-L38

{# In TSQL, subquery aggregate columns need aliases #}
{# thus: a filler col name, 'filler_column' #}
count({{ column_name }}) as filler_column

So the fix is consistent with how the project already solves the same class of problem in a sibling test.

Impact

• Helps: every strict-alias dialect (T-SQL / Fabric, Snowflake under the configurations from Missing column specification when using expression_is_true test (snowflake) #822, some Spark setups, anyone wrapping the test in a subquery).
• Breaks: nothing. select 1 as filler_column from x is valid SQL on PostgreSQL, Snowflake, BigQuery, Redshift, Databricks, Spark, T-SQL, and DuckDB. The test still returns the same row count to dbt, so pass/fail semantics are unchanged.

Out of scope

This PR does not reintroduce the condition argument removed in #700. That decision stands; this is purely a portability fix for the existing macro.

Evidence

• Stale upstream bug report: Missing column specification when using expression_is_true test (snowflake) #822
• Existing downstream workaround in dbt-fabric: https://github.com/sdebruyn/dbt-fabric/blob/main/src/dbt/include/fabric/macros/dbt_package_support/dbt_utils/generic_tests/expression_is_true.sql — patches exactly the same line to "1 as col". With this PR, that override can be deleted.

Test plan

• Added integration_tests/tests/generic/expression_is_true_subquery_alias.sql, which compiles the macro and wraps its output in an outer SELECT. Without the fix this would have failed to compile on strict-alias dialects.
• Existing data_test_expression_is_true schema tests continue to pass on the supported adapter matrix.

[sdebruyn]

Human reviewed. This would simplify the dbt-fabric adapter.