Add pandas compatible output to Series.unique (rapidsai#13959)

Partially addresses rapidsai#8175
This PR makes changes to `Series.unique`, where a cupy array is returned to match `pd.Series.unique` where a numpy array is returned.

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Lawrence Mitchell (https://github.com/wence-)

URL: rapidsai#13959

python/cudf/cudf/core/series.py

@@ -2970,6 +2970,8 @@ def unique(self):
         dtype: object
         """
         res = self._column.unique()
+        if cudf.get_option("mode.pandas_compatible"):
+            return res.values
         return Series(res, name=self.name)
 
     @_cudf_nvtx_annotate

python/cudf/cudf/tests/test_series.py

@@ -2254,3 +2254,12 @@ def test_series_nlargest_nsmallest_str_error(attr):
     assert_exceptions_equal(
         getattr(gs, attr), getattr(ps, attr), ([], {"n": 1}), ([], {"n": 1})
     )
+
+
+def test_series_unique_pandas_compatibility():
+    gs = cudf.Series([10, 11, 12, 11, 10])
+    ps = gs.to_pandas()
+    with cudf.option_context("mode.pandas_compatible", True):
+        actual = gs.unique()
+    expected = ps.unique()
+    assert_eq(actual, expected)