CDN-376: The Residual Sampling Control

CDN-376 tests whether the finite-distance residual witness from CDN-375 depends on a single Monte Carlo probe cloud. The analytic scope-mass ladder is left unchanged, while the residual estimator H = RMS(delta a - T_fit x) / RMS(delta a) is recomputed using seven probe samplers: three independent Monte Carlo volume clouds, Halton low-discrepancy volume sampling, a deterministic cube grid, a Fibonacci shell, and radial shells. The result shows that the residual values shift with sampler weighting, but the scope-threshold structure survives: at H <= 0.03, axial binary passes at D/r0=7.0 for all samplers, planar ring at D/r0=10.0 for all samplers, and axial single at D/r0=40.0 except for the pure shell sampler, which gives D/r0=56.0. The Monte Carlo blob is therefore an estimator, not the physics claim.