[RomApplication] [Fast PR] Change the size allocation of phi_elemental and psi_elemental

[Marco1410]

📝 Description

This fast PR proposes to change the size allocation of phi_elemental and _psi_elemental from elem_dofs.size() to rhs_contribution.size().

This is because in transonic potential cases it is necessary to calculate the density at an upwind point for each element located in a supersonic region. Since the supersonic regions are not known a priori, and may vary during the calculation procedure, an upwind element is associated to each element mesh.

As a result, the finite element data structure needs to be enriched and here the elemental systems are extended (from number_of_degrees_of_freedom to number_of_degrees_of_freedom + 1) to contain additions to upwind nodes, only for those elements whose local mach number is greater than a certain critical_mach.

I think this is very specific for this case, but I don't think it will be a problem in a normal case and will broaden the use of the application.

🆕 Changelog

• Updated rom_residuals_utility.h

[rubenzorrilla]

Makes sense for me. Nevertheless, I let the final approval to @Rbravo555 and @SADPR.

[Rbravo555]

This looks good to me as well. Just two comments:

1. This bug will still be present in the ELEMENTAL BUILDER AND SOLVER, as the GetPhiElemental is fixed for all elements to the same size. Creating a similar fix should be done in another PR with medium priority.
2. While launching all tests in the RomApplication with this branch, I came across two failing tests. However, they are not related to this change, but actually to missing .h5 files. HDF5 files are skipped due to the .gitignore. We should fix this with a high priority in another PR @NicolasSR, probably by saving the neural network parameters in another format not included in gitignore.