feat: Add MGR strategy for thermal single-phase flow with wells and poromechanics

[victorapm]

cc @castelletto1 @tjb-ltk @frankfeifan

Adds MGR strategy and fixes a formulation bug.

The FIM Jacobian was missing well–reservoir perforation couplings. That is now fixed in this PR. FIM assembleSystem already added perforation residuals (mass and energy), but setupSystem built the matrix pattern only from DofManager. Perforation off-diagonals are not in the DofManager graph; sequential flow+wells adds them in CoupledReservoirAndWellsBase::setSparsityPattern. In RelWithDebInfo, addToRowBinarySearchUnsorted can drop those Jacobian entries.

The well energy residual was assembled without a matching Jacobian, so the first Newton update blew it up and FIM looked unstable.

PoromechanicsSolver now takes the flow solver’s sparsity pattern, the same approach PoromechanicsConformingFractures already used. For ReservoirAndWells that includes perforation couplings; for standalone flow it is the same DofManager pattern as before.

[frankfeifan]

cc @castelletto1 @tjb-ltk @frankfeifan

Would you be to help by adding a regression test? Thanks!

Thanks, Victor! I can add one based on the sequential case I made for thermoporomechanics reservoir solver.

[victorapm]

Thanks, Frank! Feel free to push to this PR

[frankfeifan]

Thanks, Frank! Feel free to push to this PR

Hey Victor @victorapm, I've added a fim case for ReservoirThermoPoroElastic and moved both fim and sequential cases to thermoPoromechanics folder. I don't see critical issue in the linear solver for the fim case, and it runs well. The log file is attached.
output_fim.log

However, I found the solutions from the fim and sequential are quite different, e.g., see deltaPressure below (left is fim and right is sequential). And I see the fim case shows stability issue at early time. Maybe it can be mitigated by tuning some parameters in xml or mgr recipe? Please feel free to take a look.

[victorapm]

Thanks Frank! Maybe it's an issue with the FIM formulation? @tjb-ltk do you know if all the Jacobian components are being computed correctly?

[tjb-ltk]

@frankfeifan @victorapm I'll run the cases and check the well behavior.