Fix the pyrochlore Kondo coupling and unused-parameter checks in FCOrtho

[aoymt]

This PR fixes two Standard-mode generator bugs found during a systematic
per-lattice review of the generated Hamiltonians (cross-checked against
the Python port). Both are confirmed against the full base-mode
integration suite (94/94 tests pass before and after; the suite contains
no pyrochlore/fcc Standard-mode cases, so no reference files change in
this repository).

Fixes

1. Pyrochlore: the Kondo exchange coupled only the sublattice-3
   itinerant site (199095c)

   The Kondo loop in StdFace_Pyrochlore passed isite + 3 as the
   itinerant-site index:

   for (isiteUC = 0; isiteUC < StdI->NsiteUC; isiteUC++) {
     StdFace_GeneralJ(StdI, StdI->J, 1, StdI->S2, isite + 3, jsite + isiteUC);
   

   so per unit cell only the sublattice-3 itinerant site was coupled —
   to all four local spins — while the itinerant sites on sublattices
   0–2 got no Kondo coupling at all. Every other multi-site lattice
   (kagome, honeycomb) pairs the two species sublattice by sublattice,
   and the standard Kondo-lattice model couples each itinerant site to
   the local spin at the same site, so this looks like a typo for
   isite + isiteUC; this PR changes it to exactly that.

   The number of generated terms is unchanged (NsiteUC pairs per
   cell), so the nintrMax estimate still holds.

   ⚠ Behavior change: for model = "Kondo" / "KondoGC" on
   lattice = "pyrochlore", the site pairs written to
   coulombinter.def / exchange.def / hund.def change from
   (7,0) (7,1) (7,2) (7,3)-style rows to (4,0) (5,1) (6,2) (7,3)
   (per cell). Existing pyrochlore Kondo results produced by Standard
   mode were computed for the unintended Hamiltonian. Suggest a
   release-note entry.

2. FCOrtho: accepted-but-unwired interaction parameters are now
   rejected (01c8093)

   The fcc/face-centered-orthorhombic lattice has no third-neighbor
   bonds (all couplings wire the 0–2 and 0'–2' families only, and
   nintrMax counts 0 for the double-prime family), yet several
   double-prime parameters were accepted and then silently dropped:

   • fermion branch: t''/t0''–t2'' and V''/V0''–V2'' were
     neither consumed by StdFace_InputHopp/StdFace_InputCoulombV
     nor rejected by StdFace_NotUsed_*; J0''/J1''/J2'' were
     likewise missing from the NotUsed_J list (only the scalar J''
     was declared).
   • spin branch: J0''/J1''/J2'' were read and echoed to
     stdout via StdFace_InputSpinNN as if consumed, but never wired
     to any bond; the NotUsed list stopped at V, V0, V1, V',
     omitting V2, V0'–V2' and the whole double-prime t/V family.

   All of these are now declared not-used, following the complete
   declaration style of TriangularLattice.c (where the double-prime
   family actually exists and is either wired or rejected per model).

   ⚠ Behavior change: inputs that specified one of these parameters
   were silently ignored before (or echoed and then ignored, in the
   spin-model J'' case) and now abort with the standard
   unused-parameter error. Generated definition files are unchanged.

Notes for reviewers

• Same review effort as the earlier Standard-mode checks PR
  (fix/c-standard-mode-checks); like that PR, this branch is based on
  develop and intentionally touches only src/.
• The base-mode integration suite passes 94/94 both before and after;
  it has no pyrochlore or fcc Standard-mode cases, so no reference
  files change here. (Downstream reference suites that do cover
  pyrochlore Kondo need the three .def files above regenerated; the
  Python development line has already done so.)
• Fix 1 verified by hand with the dry binaries: a 1×1×1 pyrochlore
  Kondo run now emits NHund/NExchange = 4 with pairs
  (4,0) (5,1) (6,2) (7,3); fix 2 verified by confirming V'',
  t'' and J0'' inputs on lattice = "fcc" abort with the
  standard message while a clean run is byte-identical.

🤖 Generated with Claude Code