Remove tests of regridding with masks that involve randomness

My original thinking was to reduce the regridding tests considerably, to
end up with only 2 regidding tests:
- I'm thinking we don't need to cover the case with node_mask=None,
  because I assume that's already implicitly covered in pre-existing
  tests
- I think we could keep a single test of test_regrid_with_node_mask
  using one of the five current masks and a single regridding method
- I would remove test_regrid_with_random_node_mask, because I don't like
  randomness in tests: it can be problematic when you get a test failure
  and then can't reproduce it later. It does look like this test differs
  a bit from test_regrid_with_node_mask, in the destination coordinates
  in this random one differ from the source coordinates, whereas the
  destination and source are identical in the non-random version (if I
  understand the code correctly)... but I think that, for the purpose of
  what this is trying to test, that shouldn't matter.
- I would keep a single test of test_regrid_with_multivalued_node_mask,
  using a single regridding method. I would reduce the size considerably
  – e.g., a 3x4 array rather than 100x200 – and use a hard-coded array
  for src_data rather than the current random values

But Ryan O. pointed out that it can be good to have some extra test
coverage of regridding via ESMPy, so supported keeping more tests in
place. So I'm doing a more limited reduction:
- I'm removing `test_regrid_with_random_node_mask`, to remove this test
  with randomness and because it doesn't look to me like it adds
  significant coverage
- I'm changing `test_regrid_with_multivalued_node_mask` to use a small,
  hard-coded source array rather than a random array, both to remove
  randomness and to reduce the test time. (With this change, the added
  test time from this PR is insignificant, whereas previously, the tests
  in this PR added about 9 seconds, I _think_ mainly due to this larger
  test.)

I considered also changing test_regrid_with_node_mask to use a single
mask... I tried creating a mask that felt like a hybrid of the existing
masks, `[[1, 0, 0, 1], [0, 1, 1, 0], [1, 1, 0, 0]]`, but bilinear and
patch remapping failed with that mask. This made me think that I don't
understand this well enough to choose a single mask for this test, so I
left this parametrized test as is.

Author: billsacks

src/addon/esmpy/src/esmpy/test/test_api/test_regrid.py

@@ -171,60 +171,29 @@ def test_regrid_with_node_mask(mask, method):
     assert_array_almost_equal(dst.data[mask], -999)
 
 
-@pytest.mark.parametrize("method", NON_CONSERVATIVE_METHODS)
-def test_regrid_with_random_node_mask(method):
-    """Check regridding that masks random input data."""
-    n_rows, n_cols = 100, 200
-
-    src_data = np.random.uniform(-1.0, 1.0, size=n_rows * n_cols)
-    mask = src_data > 0.0
-
-    src = create_raster_field(
-        np.linspace(0.0, 200.0, num=n_cols, endpoint=True),
-        np.linspace(0.0, 100.0, num=n_rows, endpoint=True),
-        node_mask=mask,
-    )
-    dst = create_raster_field(
-        np.linspace(0.25, 199.25, num=n_cols - 1, endpoint=True),
-        np.linspace(0.25, 99.25, num=n_rows - 1, endpoint=True),
-        node_mask=None,
-    )
-
-    src.data[:] = src_data
-    dst.data[:] = 999.0
-
-    regrid = Regrid(
-        src,
-        dst,
-        regrid_method=method,
-        unmapped_action=UnmappedAction.IGNORE,
-        src_mask_values=(True,),
-        dst_mask_values=(True,),
-    )
-
-    assert np.all((dst.data <= 0.0) | (dst.data == 999.0))
-
-
 @pytest.mark.parametrize("method", NON_CONSERVATIVE_METHODS)
 def test_regrid_with_multivalued_node_mask(method):
     """Check regridding that masks multiple values."""
-    n_rows, n_cols = 100, 200
+    n_rows, n_cols = 4, 5
 
-    src_data = np.random.uniform(-1.0, 1.0, size=n_rows * n_cols)
+    src_data = np.array([-0.75, 0.75, 0.4, -0.25, 0.25,
+                         0.6, 0.8, -0.9, -0.3, -0.1,
+                         0.2, -0.6, 0.3, 0.7, -0.2,
+                         0.7, 0.6, 0.1, -0.7, -0.6])
 
     mask = np.zeros(src_data.size, dtype=int)
     mask[src_data < -0.5] = 1
     mask[(src_data >= -0.5) & (src_data < 0.5)] = 2
     mask[src_data >= 0.5] = 3
 
     src = create_raster_field(
-        np.linspace(0.0, 200.0, num=n_cols, endpoint=True),
-        np.linspace(0.0, 100.0, num=n_rows, endpoint=True),
+        np.linspace(0.0, 5.0, num=n_cols, endpoint=True),
+        np.linspace(0.0, 4.0, num=n_rows, endpoint=True),
         node_mask=mask,
     )
     dst = create_raster_field(
-        np.linspace(0.25, 199.25, num=n_cols - 1, endpoint=True),
-        np.linspace(0.25, 99.25, num=n_rows - 1, endpoint=True),
+        np.linspace(0.25, 4.25, num=n_cols - 1, endpoint=True),
+        np.linspace(0.25, 3.25, num=n_rows - 1, endpoint=True),
         node_mask=None,
     )