Fourier transience criterion in Lean without a triangle kernel: use |∫_B e^{i<s,x>} ds|^2

To bound sum_n P(|S_n| < eta) by an integral of 1/(1 - Re phi) (Chung–Fuchs criterion) in a general finite-dim inner product space, skip coordinate-wise Fejer/triangle kernels. Take K(x) = ∫_{B×B} cos<s-u, x> d(μ|B ⊗ μ|B) for a ball B of radius rho. Real.cos_sub + integral_prod_mul give K = (∫_B cos)^2 + (∫_B sin)^2 >= 0, and K >= (|B|/2)^2 for |x| <= 1/rho since cos >= 1/2 there. One integral_integral_swap gives E K(S_n) = ∫ Re phi(s-u)^n, using iIndepFun.charFun_map_fun_finsetSum_eq_prod (as in Mathlib's CLT file) for phi_{S_n} = phi^n. Then sum_{n<N} Re z^n <= 2/(1 - Re z) uniformly in N (geom_sum_eq, |1-z| >= 1 - Re z), so no r -> 1 limit is needed. The diagonal s = u is null via Measure.prod_apply + measure_singleton. Finish with lintegral_prod_symm + lintegral_sub_right_eq_self (add Haar has the right-invariance instance). Integrability of |v|^{-2} near 0 when d >= 3: integrableOn_ball_of_norm_le_rpow.