BENCH: fix time_geometric_selections (#4360)

* Fixes gh-3519

* There is no good reason to keep the flags in the
benchmark--this benchmark has been broken for over a year
so we should just use the appropriate kwargs where possible
and leave a `TODO` comment for core flags that we don't
have a substitute for

* better to have this running than hard failing all the time,
even if we've lost one of the parametrization dimensions

* if someone comes along with the confidence to say we
can delete the `TODO` comment, then we can also delete
some of the extra parametrization complexity of course

[skip cirrus]

benchmarks/benchmarks/selections.py

@@ -72,11 +72,9 @@ def time_geometric_selections(self,
                                   dynamic_selection,
                                   periodic_selection):
 
-        # set core flags for PBC accounting
-        MDAnalysis.core.flags['use_periodic_selections'] = periodic_selection[0]
-        MDAnalysis.core.flags['use_KDTree_routines'] = periodic_selection[1]
-
-        if hasattr(MDAnalysis.Universe, 'select_atoms'):
-            self.u.select_atoms(selection_string, updating=dynamic_selection)
-        else:
-            self.u.selectAtoms(selection_string, updating=dynamic_selection)
+        # TODO: Do we need a kwarg similar to old `use_KDTree_routines`
+        # flag? We used to benchmark that.
+        self.u.select_atoms(selection_string,
+                            updating=dynamic_selection,
+                            periodic=periodic_selection[0],
+                            )