[KIM docs] Reconcile periodic temperature-gradient formulas with the corrected Weber result

[krystophny]

What to build

Track correction of the consolidated periodic chapter and focused Mathematica companion, which still use (1+b_+) while the independently verified Weber reduction and production kernel use (1-b_+).

Implementation posture: tracking-only here because the documentation directory is currently outside a Git repository.

Acceptance criteria

• The periodic electrostatic and magnetic formulas use the corrected (1-b_+) coefficient.
• The focused Mathematica oracle uses the same corrected coefficient.
• A direct Weber quadrature distinguishes the corrected and stale expressions.
• Generated numerical oracle values are refreshed and agree with the Fortran regression test.
• The consolidated PDF builds without contradictory formulas.

Blocked by

A Git-backed home for the external documentation is required before a documentation PR can be opened.

[krystophny]

Additional reproducibility detail found while fixing #254: periodic_1d.tex (around lines 60–61 and 116–119) and mathematica/collisionless_periodic_kernels.wl (around lines 57–66) still encode the old collisionless magnetic-current factor/recurrence jB ~ z0/kPole. With magnitude-based z0, the derived parity-preserving regularized combination is z0/kAbs. These external artifacts therefore regenerate the pre-#254 oracle and should be updated as part of this documentation issue.