AMR: do not enforce 2:1 balance in normal direction

[kidder]

Proposed changes

Remove requirement that h-refinement must have 2:1 balance in the direction normal to faces.

Upgrade instructions

In input files that use AMR, you will need to add EnforceTwoToOneBalanceInNormalDirection: true to the list of Policies

Code review checklist

• The code is documented and the documentation renders correctly. Run
  make doc to generate the documentation locally into BUILD_DIR/docs/html.
  Then open index.html.
• The code follows the stylistic and code quality guidelines listed in the
  code review guide.
• The PR lists upgrade instructions and is labeled bugfix or
  new feature if appropriate.

Further comments

[kidder]

@nilsvu here is the unfinished (and untested) branch for what you asked about

[kidder]

@nilsvu this is now ready for review. It did not cause any problems when I ran the RandomAmr executables.

[nilsvu]

Looks good! Thanks for adding this.

[nilsvu]

Maybe set all these options to false? To me at least that's the expected behavior of AMR. Do we even have a use case for true?

[kidder]

I set them to true to reproduce previous behavior. Until setting them to false is more extensively tested, I'd rather keep them true for now.

[nilsvu]

Ok Im just worried that this will remain true and propagate to many input files and people will get unexpected behavior when trying AMR

[nilsdeppe]

I think in evolution problems we probably want it true. Geoffrey had issues if the grid size changed too much form one element to the next in the normal direction to the interface.

[nilsvu]

That's actually what I'm trying to prevent as well, in a domain that has differently sized blocks. For example, think of the thermal noise problem with thin layers on top of a substrate. You want to h-refine the substrate multiple times in z-direction, but each layer stays at h-refinement level 0.

Anyway, we can merge the PR like this and just have to keep in mind to experiment with this option.