[krystophny]

Brings the production fork up to PrincetonUniversity/GPEC develop 5499124, the merge of the release 1.6.0 PR (PrincetonUniversity#359), and keeps every fork layer on top. The previous upstream base was b894e76. This tree is the GPEC version used for the final neoclassical toroidal viscosity (NTV) torque comparison of the ITER benchmark case TC24.

develop cannot be force-pushed, so this PR sets its tree to the rebased stack. The branch rebase/upstream-54991242-20261029 holds the same tree as one commit per layer on 5499124:

Fork layer	After the rebase
#1, #2, #3 (CI, repository hygiene)	applied unchanged
#4 (previous synchronization)	its upstream content is replaced by 5499124; its .repo-hygiene-allow entry is kept, because upstream still ships the oversized DIII-D figure it allows
#5 field-line diagnostic	the source change is already upstream (PrincetonUniversity#354); only the fork's test remains
#6 projected radii (equil/direct.f)	applied without conflict
#7 geometry after resonant-field diagnostics (gpec/gpout.f)	applied without conflict

The diff against develop is therefore the upstream change from b894e76 to 5499124.

Testing was done on the TU Graz cluster with gfortran 12.2 and -O3 -march=x86-64-v3, on this tree together with the stacked follow-up PR that carries PrincetonUniversity#361 (coil/field.F only), not on this tree alone. Results:

• full make: all 13 executables built;
• the fork's two regression tests (test_direct_error_message, test_direct_projected_radius): pass;
• the Soloviev ideal, kinetic and resistive examples of .github/workflows/make_test.yml: ran to normal termination (no reference comparison);
• the reconstruction example: relative error 6.97e-3 against its tolerance of 2e-2, curl check passed.

The built pentrc is byte-identical (sha256 cedc729d…) to plain upstream 5499124 built with the same toolchain.

Chris&AI