Rebuild displaced mortar meshes during standalone Jacobian evaluation

[Andreas-Lepak]

Bug Description

FEProblemBase::computeJacobianTags() updates the displaced mesh without rebuilding displaced mortar segment meshes. Mortar segments can therefore retain old coordinates while quadrature points are mapped against updated parent faces.

When dynamic mortar regeneration was introduced in commit 8b1f83c7df, its source comment stated that mortar meshes would be rebuilt during “residual and Jacobian evaluation.” The implementation added the rebuild to residual evaluation, but not to the standalone Jacobian path.

Steps to Reproduce

Run a displaced 3D mortar problem using standalone Jacobian assembly, such as through automatic scaling. The stale mortar geometry can cause an error such as:

Unable to map 3D mortar quadrature point ... the analytical fallback did not find one unique in-domain QUAD4 inverse. Rebuilding displaced mortar immediately after updating the displaced mesh removes the failure.

Impact

Affected displaced-mortar simulations can fail or unnecessarily reduce the timestep. Residual and combined residual/Jacobian evaluation already perform the required update.

[Andreas-Lepak]

Noticed this after using the robust normal_projection from #33439, which illiminated prior projection failures.

[lindsayad]

The scaling Jacobian case is interesting; it's possible there is a gap there but the only one I can imagine is if we are using a function to drive displacements like through a Dirichlet boundary condition or if we are using a predictor. Are either of those the case? Because otherwise:

• compute scaling once: the mortar mesh evaluated at initial conditions should correspond to the mortar mesh evaluated at the scaling Jacobian
• compute scaling every time step: the solution vector should match the vector used to generate the mortar mesh at the final residual evaluation of the previous timestep

Anyway I'd accept building the mortar mesh in the scaling Jacobian calculation. I don't support generally building the mortar mesh during Jacobian assembly. In the vast majority of cases I would expect this to be duplicative work