[BUG] fix test_run_test_for_class logic check if `ONLY_CHANGED_MODU…

…LES` flag is `False` and all estimator dependencies are present (#6383)

This fixes an unreported bug pointed out by @yarnabrina in the test
`test_run_test_for_class`, in the check for the logic of
`run_test_for_class` returning an attribution of its test yes/no flag to
a reason.

If `ONLY_CHANGED_MODULES` flag is `False` and all estimator dependencies
are present, the reason returned should be `True_run_always`, but the
test checks incorrectly for another positive reason.

This is fixed - a bug with the test, not with `run_test_for_class` - as
it did not take into account properly the combination of
`ONLY_CHANGED_MODULES = False` and all dependencies present, which in
our CI occurs only in `test_all`.

sktime/tests/tests/test_test_utils.py

@@ -130,6 +130,9 @@ def test_run_test_for_class():
     if not dep_present:
         assert not run
         assert reason == "False_required_deps_missing"
+    elif not ONLY_CHANGED_MODULES:
+        assert run
+        assert reason == "True_run_always"
     elif run:
         assert reason in POS_REASONS
         assert reason_wdep == reason or reason_nodep == reason