chore(bigframes): fix ai multimodal tests (#16935)

For ai.if, ai.score and ai.classify, it's not guaranteed that the result
entries are not NULL.

Author: sycai

packages/bigframes/tests/system/small/bigquery/test_ai.py

@@ -329,7 +329,7 @@ def test_ai_if(session):
         max_error_ratio=0.5,
     )
 
-    assert _contains_no_nulls(result)
+    assert len(result) == len(s1)
     assert result.dtype == dtypes.BOOL_DTYPE
 
 
@@ -342,7 +342,7 @@ def test_ai_if_multi_model(session, bq_connection):
 
     result = bbq.ai.if_((df["image"], " contains an animal"))
 
-    assert _contains_no_nulls(result)
+    assert len(result) == len(df)
     assert result.dtype == dtypes.BOOL_DTYPE
 
 
@@ -351,7 +351,7 @@ def test_ai_classify(session):
 
     result = bbq.ai.classify(s, ["animal", "plant"])
 
-    assert _contains_no_nulls(result)
+    assert len(result) == len(s)
     assert result.dtype == dtypes.STRING_DTYPE
 
 
@@ -364,7 +364,7 @@ def test_ai_classify_multi_model(session, bq_connection):
 
     result = bbq.ai.classify(df["image"], ["photo", "cartoon"])
 
-    assert _contains_no_nulls(result)
+    assert len(result) == len(df)
     assert result.dtype == dtypes.STRING_DTYPE
 
 
@@ -374,7 +374,7 @@ def test_ai_score(session):
 
     result = bbq.ai.score(prompt)
 
-    assert _contains_no_nulls(result)
+    assert len(result) == len(s)
     assert result.dtype == dtypes.FLOAT_DTYPE
 
 
@@ -386,7 +386,7 @@ def test_ai_score_multi_model(session):
 
     result = bbq.ai.score(prompt)
 
-    assert _contains_no_nulls(result)
+    assert len(result) == len(df)
     assert result.dtype == dtypes.FLOAT_DTYPE