[npillardou]

Since #3836, the flow statistics aggregators assume that every target region of the flow solver is a CellElementRegion. When the flow solver also targets a SurfaceElementRegion (e.g. faults in conforming-fracture poromechanics), CompositionalMultiphaseStatistics / SinglePhaseStatistics with computeRegionStatistics="1" crash at setup:

***** Message from Fault (...):
'elementRegionsGroup' was expected to be a 'geos::CellElementRegion'.

• Frame 1: StatsAggregatorBase<...>::enableRegionStatisticsAggregation(...)

This PR restores the pre-#3836 behavior by iterating on ElementRegionBase / ElementSubRegionBase instead of CellElementRegion / CellElementSubRegion. Fractures are now included in the region statistics (and in the mesh-level aggregate). The sub-region kernel already only relies on fields and constitutive models that are available on face element sub-regions (element volume, pressure,
coupled solid, fluid, relperm).

It also fixes a precision issue in postAggregateStats (both single-phase and compositional): the inverse pore volume used to compute the average pressure and temperature was stored in a float, which made the average drift by up to ~1e-7 relative (e.g. average pressure above the max pressure). It is now a real64.

Changes

• StatisticsAggregatorBase(.hpp|Helpers.hpp): generic region / sub-region types in enableRegionStatisticsAggregation, forRegionStatistics and computeSubRegionRankStats.
• CompositionalMultiphaseStatistics* / SinglePhaseStatistics*: signatures updated accordingly; float → real64 for the inverse pore volume.
• ALM_multiphasePoromechanics_curvedFrac_smoke.xml: adds a CompositionalMultiphaseStatistics task with computeRegionStatistics="1" to cover the fracture case.

Testing

• ALM_multiphasePoromechanics_curvedFrac_smoke.xml runs to completion (crashed before this fix when adding statistics) and reports statistics for the Fracture region alongside the cell regions; the mesh-level pore volume equals the sum of the regions.
• Serial and 2-rank MPI runs give consistent fracture statistics.
• After the real64 fix, averages stay within [min, max] (e.g. uniform 10 MPa initial state: average = 10000000 Pa instead of 10000000.47 Pa).

Rebaseline

Requires a rebaseline:

• ALM_multiphasePoromechanics_curvedFrac_smoke gets new statistics groups in its restart output;
• tests that use flow statistics may see small changes in average pressure / temperature due to the float → real64 fix.

[jhuang2601]

This was reported previously.
It should be fixed in the pending PR (#4005), right? @MelReyCG

[MelReyCG]

No, this PR must come after the #4005