Certified enumeration of C. S. Rao's plane Śrī Yantra constraint system: a 134/681 feasibility partition realizing 128 distinct admissible figures (within B_plane)

Changes in v1.1.0: This version adds a certified nine-triangle figure renderer (render_overdetermined_figure.py) and the rendered over-determined figure (figure-overdetermined.pdf). Every figure vertex is validated against the frozen engine to machine precision. No change to the certified results (134 feasible / 681 infeasible, 128 distinct figures). --- Confirmatory computational record for the plane Śrī Yantra geometry formalized by C. S. Rao (1998, "Śrīyantra — A Study of Spherical and Plane Forms," Indian J. History of Science 33(3), 203–227). Computational result: within a registered axis-aligned parameter region B_plane, the 815 well-posed plane constraint subsets partition into 134 feasible and 681 infeasible subsets, with zero unresolved boxes and zero downgrades. Each feasible subset admits exactly one certified real figure; the 681 infeasible verdicts are machine-checked certificates of absence, not failed searches. Geometric result: the 134 feasible subsets realize 128 distinct admissible plane figures. The gap of six is exactly the rank-deficiency identity F8 − F9 + F16 ≡ 0 — three figures satisfy a sixth constraint automatically and are each counted three times in the five-subset census. The count is confirmed both under a figure metric and directly in parameter space (the three coincidences agree to 1e−12 – 1e−15). Method: a pre-registered two-tier design — certified affine-arithmetic Krawczyk interval-Newton enumeration (Tier 2, frozen, signed, timestamped, validated by a gate) cross-checked by independent multistart Newton (Tier 1). The cross-check reproduced the partition and, in the original run, detected a boundary over-certification that was then diagnosed, corrected at the tool level, re-frozen and re-run (documented as Erratum 01). Scope: all results are for the plane form and within B_plane; the spherical enumeration is deferred. This dataset builds on the corrected engine; the transcription errata it fixes are the subject of a separate corrigendum (submitted to IJHS). Includes the pre-registration lineage, hash-pinned GPG-signed and OpenTimestamps-stamped frozen-tool manifests, the certified census (results.csv, roots.jsonl), the Tier-1 cross-check, the distinct-figure analysis, the parameter-space verification, and the superseded first run (boundary-leak provenance). SHA256SUMS.txt and PROVENANCE.md make every artifact independently verifiable.