feat: add lod-decimate-chain example witnessing the modifier LOD contract

[TMHSDigital]

Summary

Second example of the "game asset pipeline" arc (after #75): a retro toy rocket — lathed body, ogive nose, three swept fins, porthole, all closed form — evaluated at LOD0/1/2 through the Decimate modifier via the depsgraph. Decimation/LOD is the pipeline stage where AI-generated code bakes reductions into the base mesh, stacks ratios silently, or loses silhouette-critical extremes.

Contract witnessed

1. Decimate is non-destructive, living in the depsgraph. Evaluated counts drop while obj.data keeps the closed-form counts (1,147 verts / 2,260 tris) — proven before and after evaluation; every evaluated reference released with to_mesh_clear().
2. COLLAPSE ratio is a target, not a guarantee. Each LOD's evaluated tris must land within 5% of ratio × base: measured 0.00% at 0.5 (1,130 tris) and 0.44% at 0.18 (405 tris) — bounds set deliberately tight from measured behavior.
3. Silhouette-critical dimensions survive. Bbox held within 1e-3 at every level (measured 7.7e-6); the closed-form bbox accounts for the asymmetric 3-fin span exactly.

Version witness: the Decimate API (decimate_type='COLLAPSE', ratio) is stable between 4.5 LTS and 5.1 — the example runs identically on both, measured values equal to the digit (verified locally on both binaries).

Falsification evidence (temporary breaks in scratch copies, 5.1.2, restored after measurement):

Probe	Break	Exit	Measured
A	second Decimate stacked at 0.5	8	tris 565 deviates 50.0% from ratio target 1,130
B	modifier dropped entirely	6	evaluated tris 2,260 == base — reduction absent
C	aggressive ratio 0.02	9	bbox drift 0.0206 — nose tip collapses (silhouette loss)

Probe C's finding is documented in the README: silhouette loss is the real failure mode of aggressive LODs, which is exactly what the bbox bound catches.

Visual / contact-sheet gate: lod-decimate-chain-contact-sheet.webp — candidate beside the pinned set (armature-bend, damped-track-aim, bmesh-gear). Per-criterion verdicts: stage darkness ✓, wedge warmth ✓, subject fill ✓ (comparable to the tubes), saturation ✓ (cream/red/teal reads saturated), thumbnail legibility ✓ (LOD progression + plaques read at 640×360). Mean luminance: candidate 139.2 vs calibration 139.4 / 100.1 / 143.4 (sRGB luma) — essentially identical to armature-bend.

Proven by live run vs inspection:

• Live run, local: check-only exit 0 on .scratch/blender-5.1.2-windows-x64/blender.exe (Blender 5.1.2, hash ec6e62d40fa9) and .scratch/blender-4.5.11-windows-x64/blender.exe (Blender 4.5.11 LTS, hash 4db51e9d1e1e) — identical measured values. Render path exit 0 on both binaries (EEVEE), pixels iterated on this machine (4 drafts: framing clips, plaque legibility fixes). All three probes live on 5.1.2.
• Live run, CI: both smoke series (4.5 LTS, 5.1) on the PR head SHA — see checks below.
• Inspection only: nothing claims otherwise.

Test plan

• Validate (structure/frontmatter), validate-manifest, validate-counts (README substring 25 examples)
• Blender 5.1 smoke — new step runs examples/lod-decimate-chain/lod_decimate_chain.py -- check-only
• Blender 4.5 smoke — same
• Ecosystem drift check; Socket Security checks
• Local binaries used for evidence: .scratch/blender-5.1.2-windows-x64/blender.exe (Blender 5.1.2) and .scratch/blender-4.5.11-windows-x64/blender.exe (Blender 4.5.11 LTS)

Integration: example dir + README, gallery.json entry, plugin manifest, smoke step, top-level README ("Game asset pipeline" category now 2, counts 24→25), generated gallery pages (read character-by-character: card alt and witnesses callout verified full), ROADMAP shipped entry.

🤖 Generated with Kimi Code