Need to update calculation for "nbp" surface_net_downward_mass_flux_of_carbon_dioxide_expressed_as_carbon_due_to_all_land_processes

[rbeucher]

Based on comments here https://docs.google.com/spreadsheets/d/1VTKrG9XcwPOVD7i-XWE2vbphm_QvSk61/edit?disco=AAABnyrkHFs

If no land use change and prescribed CO2 run then fld_s03i262-fld_s03i293. If land use change than all subtract wood respiration: fld_s03i907+fld_s03i908+fld_s03i909, summed over tiles, multiplied by land fraction and with unit conversion (divide by 1000606024365.25). If emissions driven run use fld_s03i100 but multiply by 12/44 to convert from kgCO2 to kgC

[rbeucher]

Partial progress on this in PR #333.

What has been done: The land-use-change (LUC) case for nbp — which combines NPP minus soil respiration minus wood product respiration from tiles — is now correctly divided by land fraction (fixing a separate bug where it was being multiplied instead, see #115). This covers the most common use case.

What still needs to be done: The three scenarios described in the issue are not all fully handled yet:

1. ✅ No land use change / prescribed CO2: fld_s03i262 - fld_s03i293 ÷ land fraction — this works correctly through the current formula (the tile wood respiration terms sum to zero)
2. ✅ Land use change: the full formula with wood respiration tile sums (fld_s03i907/908/909) + unit conversion ÷ land fraction — now fixed
3. ❌ Emissions-driven run: should use fld_s03i100 × (12/44) to convert from kg CO₂ to kg C — not yet implemented

The emissions-driven case requires either a separate mapping entry or a conditional derivation function that picks the right input variable at runtime depending on which variables are present in the dataset.