test: check expected columns by name instead of strict shape[1] == 5

``test_get_dataset_annotations`` asserted ``res.shape[1] == 5`` against
the gemma-rest ``/datasets/{id}/annotations`` response. gemmapy commit
``fb7afb8`` ("add evidence code to annotation processing") added
``evidence_code`` to the DataFrame projection but didn't update this
test, so the assertion has been one column behind (now 6 cols, asserts
== 5).

Replace the bare shape[1] check with an expected-column-name subset
assertion. Verifies the columns we care about are present without
breaking the next time gemma-rest emits an additive field, which is the
contract direction (server-side additive fields are backwards-compatible
for clients that don't strict-check column counts.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
EOF
)

Author: ppavlidis

tests/test_basic.py

@@ -104,7 +104,14 @@ def test_search_annotations():
 def test_get_dataset_annotations():
     res = api.get_dataset_annotations(1)
     assert type(res) is pd.core.frame.DataFrame
-    assert res.shape[1] == 5
+    # Expected columns. evidence_code was added in fb7afb8 ("add evidence
+    # code to annotation processing") but the test's bare shape[1] == 5
+    # assertion wasn't updated, so it failed once gemma-rest started
+    # emitting evidenceCode. Check by name instead of count so future
+    # additive fields don't break this test the same way.
+    expected = {"class_name", "class_URI", "term_name", "term_URI",
+                "object_class", "evidence_code"}
+    assert expected.issubset(set(res.columns))
 
 def test_get_dataset_differential_expression_analyses():
     res = api.get_dataset_differential_expression_analyses(200)