[SPARK-54842][PYTHON][TESTS] Fix test_arrow_udf_chained_iii in Python-Only MacOS26

### What changes were proposed in this pull request?
attempt to fix https://github.com/apache/spark/actions/runs/20495264978/job/58904792835,

```
Caused by: org.apache.spark.api.python.PythonException: Traceback (most recent call last):
  File "/Users/runner/work/spark/spark/python/lib/pyspark.zip/pyspark/worker.py", line 3511, in main
    process()
  File "/Users/runner/work/spark/spark/python/lib/pyspark.zip/pyspark/worker.py", line 3502, in process
    serializer.dump_stream(out_iter, outfile)
  File "/Users/runner/work/spark/spark/python/lib/pyspark.zip/pyspark/sql/pandas/serializers.py", line 781, in dump_stream
    return ArrowStreamSerializer.dump_stream(self, wrap_and_init_stream(), stream)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/runner/work/spark/spark/python/lib/pyspark.zip/pyspark/sql/pandas/serializers.py", line 120, in dump_stream
    for batch in iterator:
  File "/Users/runner/work/spark/spark/python/lib/pyspark.zip/pyspark/sql/pandas/serializers.py", line 765, in wrap_and_init_stream
    for packed in iterator:
  File "/Users/runner/work/spark/spark/python/lib/pyspark.zip/pyspark/worker.py", line 2954, in func
    for result_batch, result_type in result_iter:
  File "/Users/runner/work/spark/spark/python/pyspark/sql/tests/arrow/test_arrow_udf_scalar.py", line 930, in <lambda>
    lambda it: map(lambda x: pa.compute.subtract(x, 1), it),
                             ^^^^^^^^^^
AttributeError: module 'pyarrow' has no attribute 'compute'
```

This test passed before on macos26;
The parity test on spark connect pass;

### Why are the changes needed?

I suspect there is a cloudpickle pitfall when dealing with complicated nested lambdas, I remember I resolved a similar issue by changing the import.

### Does this PR introduce _any_ user-facing change?
no, test-only

### How was this patch tested?
cannot reproduce this issue locally, will monitor the CI

### Was this patch authored or co-authored using generative AI tooling?
no

Closes #53607 from zhengruifeng/fix_test_arrow_udf_chained_iii.

Authored-by: Ruifeng Zheng <[email protected]>
Signed-off-by: Ruifeng Zheng <[email protected]>

Author: zhengruifeng

python/pyspark/sql/tests/arrow/test_arrow_udf_scalar.py

@@ -916,26 +916,27 @@ def test_arrow_udf_chained_ii(self):
 
     def test_arrow_udf_chained_iii(self):
         import pyarrow as pa
+        import pyarrow.compute as pc
 
-        scalar_f = arrow_udf(lambda x: pa.compute.add(x, 1), LongType())
-        scalar_g = arrow_udf(lambda x: pa.compute.subtract(x, 1), LongType())
-        scalar_m = arrow_udf(lambda x, y: pa.compute.multiply(x, y), LongType())
+        scalar_f = arrow_udf(lambda x: pc.add(x, 1), LongType())
+        scalar_g = arrow_udf(lambda x: pc.subtract(x, 1), LongType())
+        scalar_m = arrow_udf(lambda x, y: pc.multiply(x, y), LongType())
 
         iter_f = arrow_udf(
-            lambda it: map(lambda x: pa.compute.add(x, 1), it),
+            lambda it: map(lambda x: pc.add(x, 1), it),
             LongType(),
             ArrowUDFType.SCALAR_ITER,
         )
         iter_g = arrow_udf(
-            lambda it: map(lambda x: pa.compute.subtract(x, 1), it),
+            lambda it: map(lambda x: pc.subtract(x, 1), it),
             LongType(),
             ArrowUDFType.SCALAR_ITER,
         )
 
         @arrow_udf(LongType())
         def iter_m(it: Iterator[Tuple[pa.Array, pa.Array]]) -> Iterator[pa.Array]:
             for a, b in it:
-                yield pa.compute.multiply(a, b)
+                yield pc.multiply(a, b)
 
         df = self.spark.range(10)
         expected = df.select(((F.col("id") + 1) * (F.col("id") - 1)).alias("res")).collect()