Depending on the order in which tests were run, the new Tool conflict

tests in test_registry would fail. One failure had to do with the

presence or lack of Tool objects in the database, and the other had to

do with an assertLogs() failing to catch logged warnings.
There were three issues behind these. One was a failure in the test, and

two were state consistency issues:


The new unit test needed Tool objects but failed to use the proper

   fixture. This wasn't caught initially due to the next issue.


Tool objects were sometimes populated when not needed by a unit

   test, and could be populated at different levels of test setup,

   causing state inconsistency depending on which tests were run first.

   The solution to this is to avoid automatically populating Tool

   objects in SCMToolRegistry when running unit tests, allowing the

   previous fixture behavior to work as it did before, and to populate

   at the right times.


A unit test in test_review.py globally disabled warnings. I can't

   determine why it did this, but it was likely for very old historical

   reasons. This affected subsequent tests. This code has now been

   removed.

Tested full test runs and various combinations of suites. All the ones

that failed before pass now.