[npillardou]

Summary

Running a case that combines a SurfaceGenerator with wells crashes during the fracturing step:

SurfaceGenerator::findFracturePlanes → getSubRegion
Error cause: castedChild == nullptr (…/wellRegion/elementSubRegions/wellRegionUniqueSubRegion)

Wells register their own nodes and elements in the node-to-element maps
(WellElementSubRegion::updateNodeManagerNodeToElementMap). SurfaceGenerator assumes every
entry of these maps is a 3D cell and casts it to CellElementSubRegion. That cast throws as soon
as the mesh contains a WellElementRegion.

Changes

SurfaceGenerator.cpp

• Add a small helper, getCellSubRegion, that returns nullptr for non-cell entries (e.g. well
  elements).
• Skip these entries in findFracturePlanes, mapConsistencyCheck, calculateNodeAndFaceSif
  and calculateElementForcesOnEdge.
• Never split a node that is attached only to non-3D elements.
• In calculateNodeAndFaceSif, average the elastic moduli over cell elements only. Before, the
  sum was divided by the total number of node-to-element entries, wells included.

The other CellElementSubRegion casts in this file go through the face-to-element maps, where
wells never appear, so they are left unchanged.

New integrated test: ALM_inclinedFault_ISG_well_smoke

This is ALM_inclinedFault_ISG_smoke, with the injection done through a well instead of a
pressure prescribed on a whole block:

• SinglePhaseReservoirPoromechanicsConformingFracturesALM + SinglePhaseReservoir + WellManager.
• One vertical BHP-controlled injector (55 MPa), perforated at mid-thickness of reservoir block 97.
• The reservoir blocks (96, 97) get a higher permeability (1e-13 m²) than the over- and
  underburden (1e-18 m²).
• The fault pressure is initialized to 35 MPa right after fracturing with a FieldApplicator.
  It is not prescribed, so it can evolve with the injection.

Input refactoring

• prescribedFaultPressure moved from ALM_inclinedFault_base.xml to the four cases that
  include it (ISG, ISG_DruckerPrager, ESG, ESG_friction_cohesion). The definition is
  unchanged.
• Reason: the well case must not have this Dirichlet condition on the fault pressure, and an
  included file cannot override it.

Testing

• Reproducer: without this fix, ALM_inclinedFault_ISG_well_smoke crashes at cycle 0 with
  the error above.
• With the fix: the case runs on 1 and 4 MPI ranks (~52 s serial, ~18 s on 4 ranks).
• Splitting: identical to the case without a well, both with 468 split nodes and 310 fault
  elements.
• The four existing ISG/ESG cases: logs are identical before and after the
  prescribedFaultPressure move (time steps, fracture states, ALM convergence, linear iterations
  and residuals).

Notes

• A new baseline is needed for ALM_inclinedFault_ISG_well_smoke.
• Found while setting up the test, not fixed here: in FieldApplicator, the targetRegions
  filter never matches. It compares subRegion.getParent().getName() to the region name, but the
  parent of a sub-region is the elementSubRegions group. The new test does not set
  targetRegions. The specification's objectPath already targets the fault.