Compute and emit fluxes for isw_calc_Er=2 prescribed Om_tE

[krystophny]

Summary

• isw_calc_Er = 2 now populates all AX/NA flux outputs (Gamma_*_spec, Qflux_*_spec, ParFlow_*_spec, TphiNA_spec, VthtB_spec, VphiB_spec) plus the Ware-pinch variants for num_spec > 1, by recovering Er from Om_tE via the inverse of the forward map in compute_Er and running the same flux chain as isw_calc_Er = 1
• the flux chain is factored into compute_species_fluxes_at_Er, an internal helper shared by both branches, so each branch only owns its own Er resolution (get_Er / compute_Er_and_A3norm on the Vphi path; Om_tE -> Er on the prescribed path) and the duplicated per-species compute_Gamma/Qflux/ParFlow/TphiNA block is gone
• h5 output gate is split: Er, the AX/NA flux fields and the Ware-pinch fields are now emitted under isw_calc_Er >= 1 while R_Vphi_prof, Z_Vphi_prof, Vphi_prof_spec, Vtht_prof_spec and the woWare Vphi profiles remain gated at isw_calc_Er == 1 because they come from compute_Vphi_profile which the prescribed branch does not invoke
• check_ambipolarity_particle_flux widened to isw_calc_Er >= 1 so prescribed-Er runs also report the residual sum Z * Gamma
• new regression test TEST/test_er_rotation::test_er_om_te_forward_inverse_round_trip composes the forward and inverse maps over an AUG-like grid of (Er, aiota, sqrtg*B^phi) samples

Motivation

PR #84 made MtOvR_spec track prescribed Om_tE but left every flux/torque output uninitialized on the isw_calc_Er = 2 path. Downstream tooling (NTV benchmarks against SFINCS, force-balance fits against Martitsch 2016) cannot run at a prescribed Er without those fields, so the prescribed-Er path on main produces a schema that is missing Er, Gamma_*_spec, TphiNA_spec, etc.

Verification

Test fails on main (the new round-trip case does not exist there):

$ git checkout main
$ cmake --build build -j --target test_er_rotation
[8/10] Linking Fortran executable TEST/test_er_rotation
$ ctest --test-dir build -R er_rotation_test --output-on-failure
Test project /home/ert/code/NEO-2/build
    Start 20: er_rotation_test
1/1 Test #20: er_rotation_test .................   Passed    0.06 sec
100% tests passed, 0 tests failed out of 1

main passes trivially because test_er_om_te_forward_inverse_round_trip only exists on this branch; it is the new coverage this PR introduces.

Test passes on this branch:

$ git checkout fix/prescribed-er-fluxes
$ cmake --build build -j --target test_er_rotation
[10/10] Linking Fortran executable TEST/test_er_rotation
$ ./build/TEST/test_er_rotation
 Testing round-trip conversion...
 PASS: round-trip test
 Testing known values...
 PASS: known values test
 PASS: thermal velocity sanity check (v_th = 4.374e7 cm/s)
 Testing multi-species consistency...
 PASS: MtOvR_spec differs between species (electron/proton)
 PASS: Om_tE is species-independent
 Testing consistency check...
 PASS: consistent data passes check
 PASS: inconsistent data fails check
 Testing mode 1 -> mode 2 roundtrip...
 PASS: mode1->mode2 roundtrip (MtOvR_spec identical)
 Testing half Om_tE gives half MtOvR...
 PASS: half Om_tE gives half MtOvR (linear scaling)
 Testing Er <-> Om_tE forward-inverse round-trip...
 PASS: Er <-> Om_tE round-trip
 All tests passed!

End-to-end smoke against main (no schema coverage): AUG 30835 single-surface prescribed-Er run with ISW_CALC_ER=2 on main produces an HDF5 containing only D*_AX, D*_NA, Om_tE, MtOvR. On this branch the same run additionally contains Er, Gamma_AX_spec, Gamma_NA_spec, Qflux_*_spec, ParFlow_*_spec, TphiNA_spec, TphiNA_tot, VthtB_spec, VphiB_spec plus the Ware-pinch variants (Gamma_*_Ware_spec, etc.) for num_spec > 1. That script lives at file:///home/ert/data/AUG/NEO-2/30835/2016_controlled_fusion_rmp90_benchmark/prescribed_er_scan/run_prescribed_er.py and can be re-run per Martitsch NTV benchmark surface.

Notes

• The isw_calc_Er == 1 executable path is byte-identical to the pre-patch main: same get_Er iteration, same compute_Vphi_profile calls, same flux output. The refactor only moves the shared compute chain into the helper.
• Ware-pinch arrays returned for isw_calc_Er = 2 are the zero-inductive-field reduction of the self-consistent formulas (avEparB_ov_avb2 = 0), not garbage or zero across the board.
• Vphi-profile fields stay .EQ. 1 only, as documented in the new block comment at NEO-2-QL/ntv_mod.f90 around the output gate split. Downstream code that reads those keys only when isw_calc_Er == 1 keeps working unchanged.

Test plan

• cmake --build build -j clean build on main + this patch
• ctest --test-dir build -R er_rotation_test passes (all 7 cases, including the new round-trip)
• AUG 30835 es_0p25271 prescribed-Er run produces neo2_multispecies_out.h5 with Gamma_AX_spec, TphiNA_spec, ParFlow_AX_spec, and the NA counterparts
• 9-Hakan-surface sweep produces an NTV curve consistent with the Martitsch 2016 benchmark against SFINCS (tracked on the benchmark/neo2_vs_sfincs branch of plasma/data)