Reorganize curved elixirs in a more systematic way

[efaulhaber]

Together with #889, this will close #887 (I changed the advection velocity in all remaining elixirs).

However, I also did a major rework of the curved elixirs:

• I added comments to all the elixirs that are running the same simulation as in some TreeMesh elixir, and I added comments to the test files explaining that these errors are supposed to be identical to the TreeMesh ones. So, whenever someone changes something that affects the errors of these elixirs, they (hopefully) don't just update the errors to the new ones, but actually copy the TreeMesh ones to make sure it's still giving the same results.
  I also updated examples/structured_3d_dgsem/elixir_advection_basic.jl to use the same setup as the TreeMesh variant.
• I made the rotated and parallelogram (only for LSA) elixirs actually use the same setup as the basic ones, so they have the same nodal values in both the IC and the exact solution, but test these with non-trivial metric terms. I also added comments explaining the purpose of these elixirs and comments in the test files, again stating that the same errors as with the basic example are expected.
• There was not a single curved LSA elixir with StructuredMesh3D with a non-constant IC. Therefore, I changed the nonperiodic LSA elixir to use a warped mesh with curved boundaries. I also changed it to use initial_condition_convergence_test instead of initial_condition_gauss, as I think that the Gauss IC is more appropriate for AMR examples.
• There was not a single curved Euler elixir in 3D with either StructuredMesh or P4estMesh, so I changed the nonperiodic Euler elixir to use a warped mesh with curved boundaries as well.
• Now to the P4estMesh2D. There was a nonconforming flag and a nonconforming unstructured flag. I made the unstructured one conforming, so we can test both the nonconforming and the unstructured feature isolated. The combination is still tested in the AMR unstructured flag. I also removed the nonperiodic boring Euler elixir and moved the structured nonperiodic test to the LSA extended elixir.
• For the P4estMesh3D, I made sure that the AMR unstructured curved elixir is actually using a mesh that is FSP. Again, I removed the unnecessary nonperiodic boring Euler elixir, and I added a nonconforming unstructured curved elixir to test all the features we have with Euler (there wasn't even a curved Euler elixir for P4estMesh3D before).
• I renamed some elixirs. For example, I added _flag to the elixirs that work on the waving flag mesh, so it's clear that these are curved elixirs.

[ranocha]

First of all: Your high-level description sounds good to me 👍 I will review this PR soon.

Your changes reduced our coverage since some non-periodic parts for the P4estMesh3D are not tested anymore, e.g., https://coveralls.io/builds/43261839/source?filename=src%2Fmeshes%2Fp4est_mesh.jl#L212 and https://coveralls.io/builds/43261839/source?filename=src%2Fmeshes%2Fp4est_mesh.jl#L478. Could you please fix that by re-adding tests?

[ranocha]

The changes look good to me (except the coverage mentioned above)

[efaulhaber]

Thanks for the review!

I added the non-periodic Euler source terms test again for P4estMesh3D.

[ranocha]

Thanks!

[sloede]

LGTM, thanks for increasing consistency and improving the tests!

I left two related comments about mesh dimensions. To summarize, I think there should be tests that do not all use the same number of cells in each dimension. However, this is not something that needs to block this PR from being merged.

[codecov]

Codecov Report

Merging #892 (d38f072) into main (9ee86eb) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #892      +/-   ##
==========================================
+ Coverage   92.71%   92.72%   +0.01%     
==========================================
  Files         229      233       +4     
  Lines       18947    18971      +24     
==========================================
+ Hits        17565    17589      +24     
  Misses       1382     1382              

Flag	Coverage Δ
unittests	92.72% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...st_2d_dgsem/elixir_advection_nonconforming_flag.jl	100.00% <ø> (ø)
...est_2d_dgsem/elixir_advection_unstructured_flag.jl	100.00% <ø> (ø)
..._dgsem/elixir_advection_amr_unstructured_curved.jl	100.00% <ø> (ø)
...s/p4est_3d_dgsem/elixir_advection_nonconforming.jl	100.00% <ø> (ø)
...t_3d_dgsem/elixir_advection_unstructured_curved.jl	100.00% <ø> (ø)
...uctured_2d_dgsem/elixir_advection_parallelogram.jl	92.31% <ø> (ø)
...tructured_3d_dgsem/elixir_advection_free_stream.jl	100.00% <ø> (ø)
...s/structured_3d_dgsem/elixir_euler_source_terms.jl	100.00% <ø> (ø)
...er_source_terms_nonconforming_unstructured_flag.jl	100.00% <100.00%> (ø)
..._source_terms_nonconforming_unstructured_curved.jl	100.00% <100.00%> (ø)
... and 5 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9ee86eb...d38f072. Read the comment docs.