Refactor InterfaceCoupling

[peterrum]

...so that it uses dealii::Utilities::MPI::RemotePointEvaluation.

depends on dealii/dealii#13173, dealii/dealii#13177, dealii/dealii#13178, dealii/dealii#13187, #127

[kronbichler]

This looks like a great simplification, now that we can get the data uniform. I guess we need some backward compatibility setting for deal.II 9.3 (I assume we could choose something similar as I did in #120 with a copy to double vectors, in absence of the most recent changes slated for inclusion deal.II regarding the float vectors).

[kronbichler]

This looks good to me, apart from the questions where we still need to decide.

[nfehn]

@peterrum Many thanks for these changes. This PR is a good example of what we describe in ExaDG's readme under "Philosophy" (https://github.com/exadg/exadg#philosophy).

[nfehn]

These changes should be tested "manually" by running FSI examples which use this functionality.

[peterrum]

@nfehn Thanks for the review. I'll try to integrate the suggestions later today or tomorrow!

[peterrum]

@nfehn I have applied the comments. I have run the cylinder_with_flag test case. I can run the other benchmarks tomorrow as well.

[nfehn]

In principle, one example should be enough because the algorithm in terms of interface coupling is the same for all examples. However, since the cylinder example is a 2d example, it could make sense to run a 3d example as well, e.g. pressure_wave.