Apply transfer2cpu before calling custom_quantity in PostprocessCallback

- Added transfer2cpu calls for dv_ode, du_ode, v_ode, and u_ode before passing them to custom_quantity
- This matches the pattern used in write_vtk.jl
- Users no longer need to handle GPU arrays in their custom_quantity functions

Co-authored-by: efaulhaber <44124897+efaulhaber@users.noreply.github.com>

Author: Copilot

src/callbacks/post_process.jl

@@ -257,6 +257,10 @@ function (pp::PostprocessCallback)(integrator; from_initialize=false)
             @notimeit timer() update_systems_and_nhs(v_ode, u_ode, semi, t)
         end
 
+        # Transfer to CPU if data is on the GPU. Do nothing if already on CPU.
+        dv_ode_cpu, du_ode_cpu = transfer2cpu(dv_ode, du_ode)
+        v_ode_cpu, u_ode_cpu = transfer2cpu(v_ode, u_ode)
+
         foreach_system(semi) do system
             if system isa AbstractBoundarySystem && pp.exclude_boundary
                 return
@@ -265,7 +269,8 @@ function (pp::PostprocessCallback)(integrator; from_initialize=false)
             system_index = system_indices(system, semi)
 
             for (key, f) in pp.func
-                result_ = custom_quantity(f, system, dv_ode, du_ode, v_ode, u_ode, semi, t)
+                result_ = custom_quantity(f, system, dv_ode_cpu, du_ode_cpu, v_ode_cpu,
+                                          u_ode_cpu, semi, t)
                 if result_ !== nothing
                     # Transfer to CPU if data is on the GPU. Do nothing if already on CPU.
                     result = transfer2cpu(result_)