review: sort order test unstable

[oliver-sanders]

We're getting a lot of flaky-failures for Cylc Review sort order tests, cause unknown.

[MetRonnie]

There may be a bug in the sorting? Specifically whenever "None" is in the last active time column

[oliver-sanders]

That would make sense, Python 2 allowed us to compare None to other types, e.g, strings, but Python 3 turned this into an error.

[oliver-sanders]

Presumed to be closed by #853 (obvs reopen otherwise!)

[MetRonnie]

Sadly not

[oliver-sanders]

Same test, but different error, before it was a ValueError, now it's a diff:

  duration_run_desc
 baz
 foo
-qux
 bar
+qux

[MetRonnie]

That's just a result of my PR #864 to make the test failures less cryptic (I think I understand the cause of the flakiness now and have asked one of the reviewers to have a go at fixing)