Let the vacuum potential's Fourier cutoffs exceed the plasma's

[CharlesCNorton]

Change - VmecInput.vacuum_mpol and VmecInput.vacuum_ntor set the Fourier cutoffs of NESTOR's scalar potential independently of the plasma's mpol and ntor, which they default to. The potential and the vacuum field keep the plasma's tangential grid; the boundary is zero-padded into the larger layout, nzeta has to carry the toroidal cutoff, and potvac is written at the potential's own size.

Cause - Fortran VMEC expands the vacuum potential to mf = mpol + 1, nf = ntor, so the vacuum solution is truncated at the plasma's resolution. On the cth boundary that truncation, not the singular quadrature, is the floor #628 discusses: an exact exterior Neumann problem (a line of dipoles inside the surface, whose total field on it is zero) gives a residual of 5.3e-3 relative to the external field with ntor = 4 for the potential at any mpol up to 32 and any grid up to ntheta = 160, nzeta = 288, and 4.8e-4 with the potential at ntor = 8. Against BIEST at twelve digits on the real cth vacuum field, NESTOR's rms error at mpol = 16 is 1.42e-3 with ntor = 4, 3.5e-4 at 6, 9.9e-5 at 8 and 7.3e-5 at 12.

Evidence - in the loop on cth_like_free_bdy at ftol = 1e-11 the change moves the boundary by tens of micrometres at its 0.19 percent beta and by 0.27 mm (mpol = 5) to 1.3 mm (mpol = 8) in rmax_surf at 2.1 percent beta on the wide mgrid; the axis moves by up to 76 micrometres. The numbers above were measured with the cross-term sign of #825; with the historical sign the vacuum error stays at 6.9e-3 for every cutoff, so the two changes only pay together.

Tests - VacuumCutoffsStayAboveThePlasmaResolution in vmec_indata_test (defaults, rejection below the plasma's cutoffs and of an nzeta that cannot carry vacuum_ntor, JSON round trip), VacuumPotentialCutoffsAboveThePlasmaResolution in vmec_in_memory_mgrid_test (the expanded run converges on the same grid, reports potvac at its size and stays within the vacuum resolution change of the base run), test_vacuum_potential_cutoffs in tests/test_free_boundary.py.

Scope - defaults are bit-identical to main: the vacuum Sizes equals the plasma's and the hand-off layout is unchanged. Raising the cutoffs costs only the vacuum solve, which grows with the potential's mode count.

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[CharlesCNorton]

fixing