Lower expected accuracy for dummy tests with conversions

This makes them more robust against these unnecessary CI failures:

`assert 76.9995346069336 == 77.0 ± 7.7e-05`

schunk_egu_egk_gripper_dummy/tests/test_dummy.py

@@ -76,7 +76,9 @@ def test_dummy_moves_to_absolute_position():
         }
 
         dummy.post(command)
-        assert dummy.get_actual_position() == pytest.approx(target_pos / 1000.0)
+        assert dummy.get_actual_position() == pytest.approx(
+            target_pos / 1000.0, rel=1e-3
+        )
         assert dummy.get_status_bit(bit=13) == 1  # position reached
         assert dummy.get_status_bit(bit=4) == 1  # command successfully processed
 
@@ -97,7 +99,7 @@ def test_dummy_moves_to_relative_position():
     dummy.post(command)
     after = dummy.get_actual_position()
     assert after < before  # we are decreasing
-    assert after == pytest.approx(before + target_pos / 1000.0)
+    assert after == pytest.approx(before + target_pos / 1000.0, rel=1e-3)
     assert dummy.get_status_bit(bit=13) == 1  # position reached
     assert dummy.get_status_bit(bit=4) == 1  # command successfully processed
 

schunk_egu_egk_gripper_dummy/tests/test_plc_communication.py

@@ -125,14 +125,14 @@ def test_dummy_supports_reading_target_position():
     dummy = Dummy()
     target_pos = 12300  # um
     dummy.plc_output_buffer[4:8] = bytes(struct.pack("i", target_pos))
-    assert pytest.approx(dummy.get_target_position()) == target_pos / 1000.0
+    assert pytest.approx(dummy.get_target_position(), rel=1e-3) == target_pos / 1000.0
 
 
 def test_dummy_supports_reading_target_speed():
     dummy = Dummy()
     target_speed = 55300
     dummy.plc_output_buffer[8:12] = bytes(struct.pack("i", target_speed))
-    assert pytest.approx(dummy.get_target_speed()) == target_speed / 1000.0
+    assert pytest.approx(dummy.get_target_speed(), rel=1e-3) == target_speed / 1000.0
 
 
 def test_dummy_supports_writing_actual_position():