Simplify diffusion test into single test file

[alexanderianblair]

Previously the diffusion regression test being used in Platypus solved both an MFEM and MOOSE variation of the problem, and compared the outputs via the L2 error norm, which required a MultiApp simulation and a Transfer to be called. This obfuscates the steps of the user setup of the MFEM FE problem.

This PR updates this test to consist of a single test file that uses MOOSE's VTKDiff tester to compare the output .pvtu files against reference data. It also removes the version of the diffusion problem in test/tests/outputs that was an effective duplicate of the main regression test for usage of the diffusion integrator in test/tests/kernels.

A minor fix to the registered base identifier in MFEMVariable has also been applied, to ensure MFEM variables can be added as AuxVariables to the problem as before.

Additionally, an option to specify the VTK output format has been added to the ParaView DataCollection wrapper. This is to allow an ASCII format to be used for tests with small output file sizes, which facilitates easier diagnosis of issues upon test failure and appears more robust when running MOOSE's XMLDiff tests on different systems, at the expense of test runtime.

[codecov]

Codecov Report

Attention: Patch coverage is 78.94737% with 4 lines in your changes missing coverage. Please review.

Files	Coverage Δ
src/base/PlatypusApp.C	72.00% <100.00%> (+0.57%)	⬆️
src/main.C	100.00% <100.00%> (ø)
src/outputs/MFEMParaViewDataCollection.C	100.00% <100.00%> (+5.55%)	⬆️
src/variables/MFEMVariable.C	92.85% <100.00%> (ø)
src/problem/MFEMProblem.C	83.27% <60.00%> (-1.75%)	⬇️

... and 6 files with indirect coverage changes

[alexanderianblair]

Note - code coverage is erroneously low for this PR as the CI is skipping the new diffusion test, since xmltodict and deepdiff need to be available for MOOSE's XMLDiff tester to work. These are added to the platypus-deps docker build in this same PR.