TST: SphericalVoronoi tests not sensitive to regions internal type

[tylerjereddy]

The regions attribute of SphericalVoronoi is defined (and currently is) a "list of lists," just like the attribute sharing the same name in the more general Voronoi class. I noticed while working on gh-19172 last night that I could simply change this to a "list of arrays" and the full SciPy test suite would pass! This isn't desirable, since some folks may depend on that structure I suppose, and the ragged data structure doesn't gain much advantage from using arrays internally in either of those classes.

Here's a sample diff that changes the return nested type without consequence:

diff --git a/scipy/spatial/_spherical_voronoi.py b/scipy/spatial/_spherical_voronoi.py
index 6b9ba8242..1f2f8e34e 100644
--- a/scipy/spatial/_spherical_voronoi.py
+++ b/scipy/spatial/_spherical_voronoi.py
@@ -224,7 +224,7 @@ class SphericalVoronoi:
         # intervals will have shape: (N+1,)
         intervals = np.cumsum(np.bincount(point_indices + 1))
         # split flattened groups to get nested list of unsorted regions
-        groups = [list(flattened_groups[intervals[i]:intervals[i + 1]])
+        groups = [flattened_groups[intervals[i]:intervals[i + 1]]
                   for i in range(len(intervals) - 1)]
         self.regions = groups

Here's a sample test modification that catches the problem:

--- a/scipy/spatial/tests/test_spherical_voronoi.py
+++ b/scipy/spatial/tests/test_spherical_voronoi.py
@@ -348,6 +348,7 @@ class TestSphericalVoronoi:
         # Tests that region integer type does not change
         # See Issue #13412
         sv = SphericalVoronoi(self.points)
+        assert isinstance(sv.regions[0], list)
         dtype = type(sv.regions[0][0])
         sv.sort_vertices_of_regions()
         assert type(sv.regions[0][0]) == dtype

Perhaps I'll label this "good first issue" and suggest making the test a little more robust than just checking a single element/region.