Periodic mesh bugfix and test

[jamiebramwell]

This ensures that the mesh nodal grid function uses discontinuous basis functions when the supplied mesh is periodic. It also fixes a bug that the nodal grid function could be double deleted between the mfem::Mesh and the StateManager.

[codecov-commenter]

Codecov Report

Merging #884 (f6dbd0f) into develop (015681c) will increase coverage by 0.05%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop     #884      +/-   ##
===========================================
+ Coverage    94.70%   94.75%   +0.05%     
===========================================
  Files          152      153       +1     
  Lines        10696    10763      +67     
===========================================
+ Hits         10130    10199      +69     
+ Misses         566      564       -2     

Impacted Files	Coverage Δ
src/serac/physics/state/state_manager.cpp	93.33% <100.00%> (+0.52%)	⬆️
src/serac/physics/tests/solid_periodic.cpp	100.00% <100.00%> (ø)
src/serac/numerics/functional/functional.hpp	100.00% <0.00%> (ø)
src/serac/physics/state/finite_element_vector.cpp	77.77% <0.00%> (+2.77%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[samuelpmishLLNL]

Can't GetNodes() return NULL for other reasons (unrelated to periodicity of the mesh)?

I've asked mfem to clarify the preconditions of this function before, but it's still essentially undocumented.

[jamiebramwell]

Yes. However, if it is a periodic mesh this will be non-null as it needs the discontinuous basis function. That is why I have the default of false on line 239.

[samuelpmishLLNL]

Ah, I misunderstood what this was doing

[samuelpmishLLNL]

Some other thoughts:

1. I'm not convinced that periodic meshes in mfem work reliably, as discussed here: Clarify limitations of periodic meshes mfem/web#211

To summarize, even the example meshes featured on mfem website will not enforce periodicity correctly in many use cases. I think it's worth managing expectations by warning the user (if we detect a periodic mesh) that things may not work, for reasons out of our control.

2. It's not obvious to me what mfem is doing regarding periodicity, so if we want to eventually support that as an official feature, we should do the due diligence of verifying that it works for different combinations of test/trial spaces.

[jamiebramwell]

Some other thoughts:

1. I'm not convinced that periodic meshes in mfem work reliably, as discussed here: [Clarify limitations of periodic meshes mfem/web#211](https://github.com/mfem/web/issues/211)

To summarize, even the example meshes featured on mfem website will not enforce periodicity correctly in many use cases. I think it's worth managing expectations by warning the user (if we detect a periodic mesh) that things may not work, for reasons out of our control.

2. It's not obvious to me what mfem is doing regarding periodicity, so if we want to eventually support that as an official feature, we should do the due diligence of verifying that it works for different combinations of test/trial spaces.

Good point. I'll add a "buyer beware" warning if a user supplies a periodic mesh.

[white238]

Can these two duplicate sections be unified into a function w/o too much effort?

[barreracruz]

I double checked in my branch with a representative test and these changes seem to fix the issues I had. Thank you!