fix(kim): align collisionless ion response with FP

[marjohma]

Summary

• add the collisionless-ion response path while retaining the Fokker--Planck electron response
• use a causal complex pole and consistent collisionless susceptibility prefactors
• make collisionless tapering and m_phi handling consistent with the FP assembly
• apply the local F0 term to every pair of overlapping hat functions, abs(l-lp) <= 1
• add ion-only FP collision-frequency scaling and kernel diagnostics
• add focused collisionless prefactor and assembly regression tests

Root cause

The collisionless charge assembly evaluated the local Debye/polarization F0
contribution only for l == lp. Linear finite-element basis functions also
overlap for nearest neighbours, so the first off-diagonal entries were omitted.
Those missing entries produced the approximately 1.6 amplitude discrepancy in
the ion-only m=7 potential even though the FP susceptibilities were already
close to their collisionless limits.

Result

After restoring the FP overlap condition, the ion-only m=7 peak-amplitude
ratio is 1.000, with a relative profile L2 difference of 10.2%. For m=6,
close to E_r=0, the peak-amplitude ratio is 1.021 and the relative L2
difference is 28.8%.

Validation

• KIM executable builds on the latest main
• collisionless prefactor and assembly tests pass
• KIM diagnostics and sparse-solver tests pass
• 38 Python benchmark tests pass
• ion-only FP and collisionless m=6/m=7 cases were rerun successfully