docs: add functional vs unit test scope rules to AGENTS.md

The split was conveyed only through examples (functional tests assert
row counts; macro unit tests use assert_sql_equal), and a contributor
could reasonably put specific-warning or compiled-SQL assertions in a
functional test without tripping any review signal. Adds three short
sentences stating the principle (functional = outcomes, unit = specifics),
the litmus (would the assertion survive a future release that ships the
same behavior via different SQL?), and a note that trivial macro
additions do not need a test.

Co-authored-by: Isaac

Author: sd-db

AGENTS.md

@@ -85,6 +85,12 @@ hatch run pytest path/to/test_file.py::TestClass::test_method -v
    - Require live Databricks workspace
    - Run with: `hatch run cluster-e2e` (or `uc-cluster-e2e`, `sqlw-e2e`)
 
+### What to Assert in Each Test Type
+
+Functional tests assert user-visible outcomes: the model materializes and the resulting rows are correct. Unit and macro tests assert implementation details like the exact SQL a macro generates or the exact text of a warning.
+
+When there is nothing meaningful to assert in either category, the change does not need a test.
+
 ### Test Environments
 
 - **HMS Cluster** (`databricks_cluster`): Legacy Hive Metastore