Fix databse ordering assumptions in schedule tests

[drnlm]

The schedule tests (mostly in the json and ical tests) make some assumptions about how the data will be returned from the database (explicitly mentioned in test_json_view_admin).

Recent jobs have being failing on postgres in an inconsistent way (failing tests may pass if rerun, etc) . My current guess is that this is because we're now testing against postgres 15, instead of 14 as we would have been last year (at least it's the only change in the setup that looks relevant I can see) so we need to revisit these assumptions.