Norne benchmark case

Source: the Norne field simulation model published by Equinor and the Norne
licence partners through NTNU and the Open Porous Media (OPM) initiative,
https://github.com/OPM/opm-data (folder "norne", deck NORNE_ATW2013.DATA).

Licence: Open Database License (ODbL) 1.0, https://opendatacommons.org/licenses/odbl/1-0/

This folder is a derived database. Grid geometry, cell connections and static
properties come from the model's EGRID/INIT output. Saturations and pressures
come from running the public deck with OPM Flow 2026.04 (Python bindings) from
6 Nov 1997 to 1 Dec 2006. The only change to the deck was removing RPTSCHED
report settings from the schedule so that restart output is written monthly.

Check against published results: OPM's opm-tests repository carries Eclipse
2014.2 summary results for the same deck (norne/ECL.2014.2). At 1 Dec 2006:
  cumulative oil    -0.2%
  cumulative water  +0.4%
  cumulative gas    -0.1%
  water injected    +0.1%
  average field pressure within 0.1 bar over the run
Published results contain field and well summaries only, no 3D cell data.

Files
  model.json       dimensions, frame dates, wells (path, completions, role per frame), rates (Sm3/day, averages between frames)
  geometry.bin     float32, 8 corners x (x, y, depth) per active cell, metres, relative to model.center
  neighbours.bin   int32, 6 face neighbours per active cell (-I, +I, -J, +J, up, down), -1 = open face
  ijk.bin          uint8, 0-based I, J, K per active cell
  static.bin       float32 blocks of PORO, PERMX, PERMZ, NTG, FIPNUM, DEPTH
  dynamic.bin      per frame: SWAT uint8 (/255), SGAS uint8 (/255), PRESSURE uint16 scaled to dynamic.pressureRange
