Reorganize tests

[ranocha]

Once #569 is resolved, I would prefer to structure the tests similarly.

[ranocha]

We should also look at #372 when doing this reorganization. I did some very preliminary tests to see whether some order is preferrable.

julia> @time redirect_stdout(devnull) do # by equation
           run(`$(Base.julia_cmd()) -e 'include("examples/tree_2d_dgsem/elixir_advection_basic.jl");
                                        include("examples/tree_3d_dgsem/elixir_advection_basic.jl");
                                        include("examples/structured_2d_dgsem/elixir_advection_basic.jl");
                                        include("examples/structured_3d_dgsem/elixir_advection_basic.jl");'`)
           run(`$(Base.julia_cmd()) -e 'include("examples/tree_2d_dgsem/elixir_euler_ec.jl");
                                        include("examples/tree_3d_dgsem/elixir_euler_ec.jl");
                                        include("examples/structured_2d_dgsem/elixir_euler_ec.jl");
                                        include("examples/structured_3d_dgsem/elixir_euler_ec.jl");'`)
       end
164.508251 seconds (911.44 k allocations: 49.725 MiB, 0.00% gc time, 0.07% compilation time)

julia> @time redirect_stdout(devnull) do # by mesh
           run(`$(Base.julia_cmd()) -e 'include("examples/tree_2d_dgsem/elixir_advection_basic.jl");
                                        include("examples/tree_3d_dgsem/elixir_advection_basic.jl");
                                        include("examples/tree_2d_dgsem/elixir_euler_ec.jl");
                                        include("examples/tree_3d_dgsem/elixir_euler_ec.jl");'`)
           run(`$(Base.julia_cmd()) -e 'include("examples/structured_2d_dgsem/elixir_advection_basic.jl");
                                        include("examples/structured_3d_dgsem/elixir_advection_basic.jl");
                                        include("examples/structured_2d_dgsem/elixir_euler_ec.jl");
                                        include("examples/structured_3d_dgsem/elixir_euler_ec.jl");'`)
       end
147.843229 seconds (794.04 k allocations: 42.443 MiB, 0.00% gc time, 0.04% compilation time)

julia> @time redirect_stdout(devnull) do # by dimension
           run(`$(Base.julia_cmd()) -e 'include("examples/tree_2d_dgsem/elixir_advection_basic.jl");
                                        include("examples/tree_2d_dgsem/elixir_euler_ec.jl");
                                        include("examples/structured_2d_dgsem/elixir_advection_basic.jl");
                                        include("examples/structured_2d_dgsem/elixir_euler_ec.jl");'`)
           run(`$(Base.julia_cmd()) -e 'include("examples/tree_3d_dgsem/elixir_advection_basic.jl");
                                        include("examples/tree_3d_dgsem/elixir_euler_ec.jl");
                                        include("examples/structured_3d_dgsem/elixir_advection_basic.jl");
                                        include("examples/structured_3d_dgsem/elixir_euler_ec.jl");'`)
       end
145.176463 seconds (791.85 k allocations: 42.283 MiB, 0.00% gc time, 0.04% compilation time)

Based on these preliminary results, it looks like we shouldn't sort the tests by equations but by dimension/mesh.

[sloede]

Thanks for testing this out! If the optimal sorting matches (or is close to) the alphabetical ordering of the example folders, it would be nice to retain the same order and get extra karma for staying consistent!