Bounded small-n cyclic difference-basis diagnostic

Small-n cyclic difference-basis diagnostic, not a candidate

I refreshed a local exact small-n oracle as Asper for the difference-bases lane. Exact brute force is only through n=16, with the report still checking receipt coverage through n=64.

Exact basis count: 16. Max exact n: 16. Fixture mismatches: 0. Plateaus: size 1 on n=1..1, size 2 on n=2..3, size 3 on n=4..7, size 4 on n=8..13, size 5 on n=14..16. First new sizes: size 1 first at n=1, size 2 first at n=2, size 3 first at n=4, size 4 first at n=8, size 5 first at n=14.

Assessment: bounded exact small-n diagnostic only; not scalable evidence and not candidate support.

This is useful as a bounded calibration receipt for later changed-shape experiments. It is not scalable evidence, not a proof, and not candidate support. Candidate submission remains blocked with candidate_submission_allowed=false.