Make sure we serialize all active plugins.

[bangerth]

As @Francyrad points out in https://community.geodynamics.org/t/impossible-to-creare-an-inner-core-using-the-dynamic-core-cc-fixed-by-me-function-wrong-unit-of-measure/3910/13, the dynamic core plugin does not deal well with serialization. That's because we don't serialize it at all: None of the boundary temperature plugins are in fact serialized at all, nor do they have to functionality to be serialized. Looking around, this seems to be true for all sorts of plugin systems. For most plugins that doesn't matter all that much because they have no state, but the one @Francyrad is using does and his observation shows the issue. We should really address this somehow.

[Francyrad]

Hi, should I make a try?

[bangerth]

What's currently missing is that we need to go through all of the plugin classes and see whether they have state that needs to be serialized.

There is also this piece of code that we should address:

   template <int dim>
    void
    Visualization<dim>::save (std::map<std::string, std::string> &status_strings) const
    {
      // Serialize into a stringstream. Put the following into a code
      // block of its own to ensure the destruction of the 'oa'
      // archive triggers a flush() on the stringstream so we can
      // query the completed string below.
      std::ostringstream os;
      {
        aspect::oarchive oa (os);
        oa << (*this);
      }

      status_strings["Visualization"] = os.str();

//TODO: do something about the visualization postprocessor plugins
    }


    template <int dim>
    void
    Visualization<dim>::load (const std::map<std::string, std::string> &status_strings)
    {
      // see if something was saved
      if (status_strings.find("Visualization") != status_strings.end())
        {
          std::istringstream is (status_strings.find("Visualization")->second);
          aspect::iarchive ia (is);
          ia >> (*this);
        }

//TODO: do something about the visualization postprocessor plugins
    }