Adaptivity improvements

[lindsayad]

• Add support for adaptivity with static condensation. Follow-on from StaticCondensation can also be a DofMap #4140 and in response to Interior penalty HDG idaholab/moose#27188 (review)
• Clean refinement flags after projecting vectors. This ensures that if there is no mesh refinement before another
  call to project vectors that vector projection will work. Refs Don't do unnecessary refine in reinit idaholab/moose#30742

[moosebuild]

Job Coverage, step Generate coverage on 7d2f546 wanted to post the following:

Coverage

Inconsistent report tags were found between the head and base reports.
This can happen when reports are missing from either the head or the base.

Inconsistent tags:
distributed_odd
Full coverage report

This comment will be updated on new commits.

[lindsayad]

MetaPhysicL failure again 😦

[roystgnr]

I think I'm happy with this, though I'd love for @jwpeterson to take a look too.

[roystgnr]

Likewise for these two. Though I admit we're now up to three more arguments, and my suggestion is getting less compelling, it just feels confusing to have temporary storage saved as members.

[lindsayad]

So you would prefer this?

void StaticCondensationDofMap::add_uncondensed_dof(
    const dof_id_type full_dof_number,
    const bool involved_in_constraints,
    std::unordered_map<dof_id_type, dof_id_type> & uncondensed_global_to_local_map,
    std::unordered_set<dof_id_type> & local_uncondensed_dofs_set,
    std::unordered_map<processor_id_type, std::unordered_set<dof_id_type>> &
        nonlocal_uncondensed_dofs,
    std::vector<dof_id_type> & elem_uncondensed_dofs,
    dof_id_type & uncondensed_local_dof_number,
    std::unordered_set<dof_id_type> constraint_dofs)

[lindsayad]

c2bd8d6 uses local variables

[roystgnr]

Yeah, honestly. It feels like that "scratch space" could be members of a nested class, even?

I wouldn't go that far now, though. If this was a public API I'd be anxious over getting the best design before we exposed it to users and had to support it, but with a private API we can redo at any time I'm just expressing mild preferences.

[roystgnr]

We can merge without kicking that test - MetaPhysicL failure again. I'm so baffled by that one. LIBMESH_RUN doesn't affect MetaPhysicl make check and we don't set METAPHYSICL_RUN so we shouldn't be getting anything non-deterministic from MPI, and none of the unit tests there respect --n_threads= either.

[lindsayad]

I force pushed the last commit just with the blank line removals that @jwpeterson pointed out. I'll merge if/once this passes CI