Map region arrays onto the leaf grid for LGR (EQLNUM, PVTNUM, SWATINIT, FIP)

[hnil]

What

Region arrays (EQLNUM, PVTNUM, SWATINIT, FIPNUM and the datum-pressure regions) are read from the input grid field properties but then indexed by leaf cell. Without LGRs the leaf coincides with the input grid, so this is correct. With a CARFIN the leaf has more cells and a different ordering, and the arrays are silently misused.

Each case is mapped onto the leaf with LookUpData::assignFieldProps{Int,Double}OnLeaf, so a refined cell inherits its parent cell's value. That is the identity on an unrefined grid, so non-LGR runs are byte-identical.

Why it matters

These are wrong-answer bugs on decks upstream already supports, not new-feature enablers:

• EQLNUM — equilnum() copied the input-grid array into a leaf-sized vector with std::ranges::transform, misaligning coarse cells and leaving every refined cell in equil region 1. Refined cells then equilibrate against the wrong OWC, so the refined region initialises at the wrong water saturation.
• PVTNUM — setRegionPvtIdx() indexed the input-grid array by a leaf cell index. Out of bounds for refined cells; once a region's first cell was refined this surfaced as "table has 0 sampling points".
• SWATINIT — input-grid values copied into a leaf-indexed vector.
• FIPNUM and the FIP region arrays — createLocalRegion_ built the array from the input grid and then only resize()d it to the leaf cell count. std::vector::resize keeps the existing values and zero-pads the appended entries, so every refined leaf cell got region 0 and was dropped from the FIP/field sums. FPR is the FIPNUM-region pressure sum, so excluding the refined cells — which sit near the well, at a different pressure — biases the reported field pressure while the per-cell maps stay correct.
• Datum-region pressures — the PPO/PPG/PPW and RPP* keywords build a RegionPhasePoreVolAverage from PVTNUM/FIP arrays and index it by leaf cell, reading past the end of the input-grid arrays.

The last commit is a guard rather than a fix: the solvent/polymer/biofilm/MICP initial conditions (SSOL, SPOLY, SPOLYMW, SMICR, SBIOF, SOXYG, SUREA, SCALC) are read the same way and would need the same treatment. LGR is black-oil only today, so this fails early with a clear message instead of producing silently wrong results.

How to reproduce

flow opm-tests/lgr/SPE1CASE1_CARFIN.DATA

Compare FPR and the initial SWAT in the refined region against the equivalent unrefined run. This PR changes simulation output on LGR decks — that is the point; the previous values were wrong. Non-LGR output is unchanged.

The quantified effect was measured on an internal two-phase LGR model (not in opm-tests): FPR absolute error against the reference dropped from 9.06 to 0.055, matching the non-LGR baseline of ~0.12, serially and at np=2; initial SWAT moved from a gross WOC error to the same ~2% per-cell spread as the non-LGR case.

Implementation notes

• Uses only existing opm-grid API — LookUpData / assignFieldPropsIntOnLeaf (opm/grid/LookUpData.hh). No grid-side change is needed.
• equilnum() now takes the GridView, which LookUpData requires.
• In OutputBlackoilModule, the mapped arrays are captured by value in the lookup lambda so they outlive the constructor; decltype(auto) is kept for the required reference-to-vector semantics.
• needsTranslation=false throughout, preserving the previous raw get_int values.

Testing

Built and run against opm-common and opm-grid master. Non-LGR regression cases unchanged.