Restore global state modified during tests

Specifically, numpy's error handling settings and yt's logging level.

yt/data_objects/tests/test_image_array.py

@@ -9,14 +9,22 @@
 from yt.data_objects.image_array import ImageArray
 from yt.testing import requires_module
 
+old_settings = None
+
 
 def setup():
+    global old_settings
     from yt.config import ytcfg
 
     ytcfg["yt", "internals", "within_testing"] = True
+    old_settings = np.geterr()
     np.seterr(all="ignore")
 
 
+def teardown():
+    np.seterr(**old_settings)
+
+
 def dummy_image(kstep, nlayers):
     im = np.zeros([64, 128, nlayers])
     for i in range(im.shape[0]):

yt/utilities/lib/tests/test_alt_ray_tracers.py

@@ -5,12 +5,13 @@
 from yt.testing import amrspace
 from yt.utilities.lib.alt_ray_tracers import _cyl2cart, cylindrical_ray_trace
 
-left_grid = right_grid = amr_levels = center_grid = data = None
+left_grid = right_grid = amr_levels = center_grid = data = old_settings = None
 
 
 def setup():
     # set up some sample cylindrical grid data, radiating out from center
-    global left_grid, right_grid, amr_levels, center_grid, data
+    global left_grid, right_grid, amr_levels, center_grid, data, old_settings
+    old_settings = np.geterr()
     np.seterr(all="ignore")
     l1, r1, lvl1 = amrspace([0.0, 1.0, 0.0, -1.0, 0.0, 2 * np.pi], levels=(7, 7, 0))
     l2, r2, lvl2 = amrspace([0.0, 1.0, 0.0, 1.0, 0.0, 2 * np.pi], levels=(7, 7, 0))
@@ -21,6 +22,10 @@ def setup():
     data = np.cos(np.sqrt(np.sum(center_grid[:, :2] ** 2, axis=1))) ** 2  # cos^2
 
 
+def teardown():
+    np.seterr(**old_settings)
+
+
 point_pairs = np.array(
     [
         # p1               p2

yt/utilities/tests/test_set_log_level.py

@@ -2,6 +2,19 @@
 
 from yt.utilities.logger import set_log_level
 
+old_level = None
+
+
+def setup():
+    global old_level
+    from yt.utilities.logger import ytLogger
+
+    old_level = ytLogger.level
+
+
+def teardown():
+    set_log_level(old_level)
+
 
 def test_valid_level():
     # test a subset of valid entries to cover