feat(scaffold): post-mount stem for plinth-hosted readouts (#286)

[bradleypmartin]

Closes #286 once headset smoke confirms the post reads as plinth hardware.

Summary

• Surfaced from PR feat(scaffold): extrude readout back-plate to a depth slab (#270) #285 smoke (third round on feat(scaffold): display-mock visual-language pass on plinth-mounted readouts (post-#252) #270's depth-slab): readout slab still reads as a "floating monitor that happens to be near the plinth," not as something physically mounted on it.
• Adds a thin cylindrical ReadoutPost primitive (src/scaffold/ui/ReadoutPost.ts) registered in its own 'surface'-oriented plinth slot at the same XY as each readout. Post is statically anchored to the working surface (no faceCamera overwrite); the readout slot's localXYZ[2] lifts the readout group by READOUT_POST_LENGTH along the surface normal so the readout group origin sits at the post tip.
• PanelReadout.faceCamera is unchanged — today's world-Y yaw billboard is preserved across all four cluster scenes. No changes to back-plate, text layout, dimensions, or rotation math.
• Plan doc: _private/plans/286-readout-post-mount.md v3 (post-roundtable + second-Sonnet sanity pass).

Architecture context

v1 of the plan proposed yawing the panel around the post axis. Three-vendor /roundtable-plan-review returned overall HOLD — four convergent HIGH findings on broken rotation math. v2 abandoned that design in favor of two slots per readout (post + panel as separate plinth-registered primitives, panel preserves today's billboard). v2 → second-Sonnet PROCEED with two LOW tightening nits, folded into v3 (explicit orientation: 'surface' at the post slot sites, plus a wiring test asserting the post quaternion equals R_x(-PLINTH_TILT_DEFAULT) after createPlinth placement).

The post is purely visual signaling that the readout is mounted on the plinth — not a kinematic element. Panel center orbits the lifted slot point on an 8 mm radius circle as yaw varies (panel-back is 8 mm behind group origin in local Z); at panel scale (half-widths 0.20–0.38 m) this is visually invisible.

Test plan

• npx vitest run — 46 files, 610 tests passing (6 new tests in test/scaffold/ui/ReadoutPost.test.ts).
• npx tsc --noEmit — clean.
• pre-commit — gitleaks + eslint + standard fixers clean.
• Headset smoke on Cloudflare PR preview: readout reads as mounted on a stem extruded from the plinth surface (kills the "floating monitor" read from feat(scaffold): extrude readout back-plate to a depth slab (#270) #285). Panel yaw behavior unchanged. Post stays anchored to the surface from any viewing angle. Quest 3S 72Hz hold preserved.
• Pancake smoke on Cloudflare PR preview: default OrbitControls pose still frames plinth + math object + readout acceptably in all four scenes (blocking acceptance criterion per plan §5; if broken, retune PLINTH_*_READOUT_Y per-scene in this PR before merge).
• SceneRack scene-switch cycle: no flicker, post re-mounts cleanly across hops.

Refs: #286 (this), #285 (smoke source), #270 (closed; chrome family), #225 (parent sub-epic), #221 (v1.0 staging epic), #271 (composition partner).

[github-actions]

Preview deploy ready — commit 0bf09e819fbde574460a487854234e8fa6ce6223

https://geometer-pr-287.1bradley-martin1.workers.dev

Open this URL in the Quest browser to smoke-test in headset.