Use the test ID as the unique component of the test run collection

This is a reasonable thing to do in general, but is important when
using pytest-randomly because when using that plugin without the
--randomly-dont-reset-seed flag, the seed is forced to the global
value before every test is run. This leads to makeTestCollection
returning the same random integer for every test and the runs
clash in the shared repo.

tests/test_calibrate.py

@@ -109,7 +109,7 @@ def tearDownClass(cls):
 
     def setUp(self):
         super().setUp()
-        self.butler = butlerTests.makeTestCollection(self.repo)
+        self.butler = butlerTests.makeTestCollection(self.repo, uniqueId=self.id())
 
         self.dataId = {"instrument": "notACam", "visit": 101, "detector": 42}
         # CalibrateTask absolutely requires an ExpandedDataCoordinate