fix(KIM): apply ion FLR scale consistently

[krystophny]

Summary

• make ion_flr_scale_factor affect only positively charged ion species, never electrons
• apply the ion-only factor consistently to all six FP analytic prefactors proportional to a signed gyroradius (1/Omega_sigma)
• consume stored rho_L_cc exactly once in global integration geometry
• add a focused regression covering electron invariance, single ion scaling, and the production geometry seam

Derivation convention

The finite-collision derivation gives all four response channels a common 1/Omega_sigma external factor. The artificial ion-FLR knob therefore applies to each ion channel. Separately, background construction already stores an effective scaled ion rho_L, so real-space Bessel geometry must not multiply that stored radius again.

Red/green evidence

Before the fix, changing the ion scale from 1 to 4 multiplied the manufactured electron FP prefactors by 4. The geometry seam did not exist, and the global path directly multiplied an already-scaled rho_L_cc by the knob again. The new test now covers all six affected analytic prefactors and the one-time geometry rule.

Verification

• test_ion_flr_scale_factor
• test_integration_methods
• test_flr2_fourier_kernel
• test_periodic_background_build
• test_periodic_assembly
• test_kim_solver_em
• git diff --check

All passed. Independent re-review: no remaining findings.

Closes #257

[marjohma]

Closing this draft so it can be recreated under the requested marjohma PR authorship. The replacement will use the same branch and commit.

[marjohma]

Replacement: #280, opened by marjohma from the same branch and commit.