Make Taal pass Taam tests

[sloede]

The idea for this PR is to make Taal pass all Taam tests. That is, we use the same number of analysis nodes in Taal as for Taam, and we temporarily switch to the old time step calculation scheme. Then, we confirm that Taal passes all Taam tests by exactly copying the Taam error norms.

Once we created elixirs for all Taam tests, i.e., formally have at least the same level of coverage (feature-wise) as we did before, we can decide to switch back to the new time step calculation scheme (and possibly also change the number of analysis nodes again).

While this might result in more work (since we have to re-run all tests in the end to get new reference values for Taal after we switch the time step calculation), it gives us the guarantee that Taal produces exactly (up to machine precision) the same results as for Taam for all tests.

To ensure that no test is accidentally left behind, all @testset names in Taal and Taam are prefixed by taal-check-me. If, for a given test, the reference values from Taam are copied to the Taal test file and it is confirmed that the results match, this should be changed to taal-confirmed. In the end, we should only have tests with taal-confirmed in all test files (at least for the parameter file/elixir tests), and can delete it when we switch the time step calculation scheme.

Notes
Code that needs to be updated/adapted after all tests are synchronized is marked by

# FIXME Taal restore after Taam sync

TODO

• Port fixes to blob/KHI tests from [OUTDATED][WIP]: Make Taal and Taam produce identical (to floating point accuracy) results #253
• What to do about the index error in MHD (37074b8)? Maybe:
  • Fix in master
  • merge master to dev
  • merge dev to here?
    In that case: Also fix calcflux (see below), since references will change anyways...
• Somehow, even though I (@sloede) checked in manually before, 1D advection amr nonperiodic is broken. However, while the reference values also do not match the Taam results, Taal and Taam produce identical (FP) output?!
• In 2D advection, currently Taal still uses the new time step calculation - but the results still match? Fixed in d6ad86f
• Convert examples/3d/elixir_euler_blob_amr.jl once PositivityPreservingLimiterZhangShu #274 is merged.

After action cleanup

• Apply fixes to MHD's calcflux (c6f4cf4), this will need done for 1D, 2D, and 3D
• Fix calculation of c_h for MHD

[ranocha]

In 2D advection, currently Taal still uses the new time step calculation - but the results still match? Needs investigating...

These use constant velocities like (1.0, 1.0). Hence, the new time step calculation in Taal is equivalent to the old one in Taam if the CFL number is scaled by a factor of two.

[sloede]

In 2D advection, currently Taal still uses the new time step calculation - but the results still match? Needs investigating...

These use constant velocities like (1.0, 1.0). Hence, the new time step calculation in Taal is equivalent to the old one in Taam if the CFL number is scaled by a factor of two.

Thanks, this was the case - I had forgotten to update the CFLs. Fixed in d6ad86f

[ranocha]

I've ported the new AMR controller from #266 to dev. You can use them by merging dev into this branch.

[ranocha]

When sorting the tests, I would like to make this test without modifications the first one and check the other time integration methods afterwards.

[sloede]

👍