fem_system_ex2 local->distributed solution

There doesn't seem to be any way to get this behavior back in reinit()
without creating a performance regression: most applications simply
never need to move data in that direction.

Refs #1593.

examples/fem_system/fem_system_ex2/fem_system_ex2.C

@@ -82,7 +82,9 @@ void setup(EquationSystems & systems,
   imms.save_initial_mesh();
 
   // Fill global solution vector from local ones
-  aux_sys.reinit();
+  aux_sys.current_local_solution->close();
+  *aux_sys.solution = *aux_sys.current_local_solution;
+  aux_sys.solution->close();
 }
 
 
@@ -119,7 +121,9 @@ void run_timestepping(EquationSystems & systems, GetPot & args)
 
       out << "Solving Solid" << std::endl;
       solid_system.solve();
-      aux_system.reinit();
+      aux_system.current_local_solution->close();
+      *aux_system.solution = *aux_system.current_local_solution;
+      aux_system.solution->close();
 
       // Carry out the adaptive mesh refinement/coarsening
       out << "Doing a reinit of the equation systems" << std::endl;