QL: single-helicity misalignment drive in ripple_solver_ArnoldiO2

[krystophny]

Implements #118. This branch is stacked on #124 and #121 and remains open against main. The separate child branch for #122 enables source-only execution when this drive supplies a nonzero source.

Summary

• add isw_hel_drive, m_theta_hel, hel_brad_re/im, and hel_phim_re/im to ntv_input
• add the single-helicity right-hand side to the non-axisymmetric Arnoldi O2 solve, default off
• add the two velocity moments needed by the drive and unload their storage with the other collision moments
• expose and test the production source-activation predicate used by the QL: source-only in_file_pert='none' feeds a zero source to the Arnoldi eigensolver (NaN, endless zhseqr failures) #122 child
• document input phase, amplitudes, and CGS units

Invariants

Switch-off behavior remains unchanged. The drive preserves the existing Arnoldi O2 discretization, collision-moment storage layout, species ordering, ABI, field-line boundary conditions, and Fourier convention. The source uses CGS/Gaussian units and the CAS-checked exp(+i(m theta + n phi)) phase map. No branch in this stack is authorized to merge before the programme-wide gate.

Verification

Test fails on main

With the source-activation regression applied before the drive implementation:

$ cmake --build --preset default --target test_ntv_hel_drive_input
Error: Symbol ‘has_helical_drive_source’ referenced at (1) not found in module ‘ntv_mod’
Error: Function ‘has_helical_drive_source’ at (1) has no IMPLICIT type
ninja: build stopped: subcommand failed.

Test passes after fix

$ fo
Static: OK
Build: OK
Tests: OK (9/9)
Lint: OK
All stages passed

The full pipeline passed twice consecutively after #124 serialized the duplicate build tests.

$ make test
100% tests passed, 0 tests failed out of 7

$ ctest --test-dir build -R "^ntv_hel_drive_input_test$" --output-on-failure
1/1 Test #...: ntv_hel_drive_input_test ... Passed

$ git diff --check
<no output>

The 16 separately managed external-data regressions remain registered under regression;external-data; this branch does not mark that unavailable-data gate passed.

[krystophny]

Additional verification: the rotating aligned-pair regression passes. With Om_tE = 3e5 rad/s (convention Om_tE = c Er/(aiota sqrtg_bctrvr_phi), i.e. the rigid toroidal ExB precession -c dPhi0/dpsi_pol), the corrugation-only drive produces a helical-harmonic current of 46.3 (solver units, electron A1 channel); adding the potential aligned in the solver's rotating frame — the geometric aligned amplitude times -n(iota*B_th + B_ph)/(iota*(m*B_ph - n*B_th)), the correction being the parallel return flow of the ExB rotation — suppresses it to 0.12 (factor ~390), while the sign-flipped control doubles it to 92.7. This exercises the drive calibration, the hatOmegaE rotation-term normalization and the sign conventions in a single test (CAS derivation and run inputs live with the physics programme).

[krystophny]

@GeorgGrassler please help Sergei to get this branch working when you have some time.