[KAMEL] Benchmark integral ions against the drift-kinetic limit

[krystophny]

What to build

Add a dual-evaluation benchmark that compares KIM's integral ion tensor with the established Heyn/Markl drift-kinetic ion tensor without changing the configured evolution model. Electrons remain on the drift-kinetic model in every benchmark case.

The benchmark must connect the symbolic reduction to numerical KIM, periodic-window, and QL-Balance results and report why each tolerance is finite.

Acceptance criteria

• Benchmark mode computes and records both old and new ion D11, D12, D21, and D22 while evolution continues to use the explicitly selected model.
• The limiting fixture uses k-perp-rho_i approaching zero, ell=0, B-parallel=0, and the local diagonal radial-wave-number limit.
• The Mathematica oracle reduces all four integral coefficients to the old formulas under that limit.
• A synthetic one-radius test agrees at algebraic precision.
• A constant-profile periodic-window test agrees point by point at the predicted discretization order.
• An end-to-end QL-Balance case writes absolute and relative old/new residuals for all four coefficients.
• Electron coefficients are unchanged from the established drift-kinetic path.
• Algebraic, DFT/quadrature, and finite-FLR asymptotic errors have separate documented tolerances.
• B-parallel=0 recovery and a nonzero B-parallel extension test are both retained.
• The limiting case is part of automated regression or golden coverage.

Blocked by

• [QL-Balance] Couple one periodic KIM mode with compact global embedding #290