Deepen cooperative learning, role handovers and people feedback

[NellInc]

Changes

• Evidence-led cooperative shift review, with direct comparison navigation.
• Optional shared packing predictions, frozen at rehearsal, compared with real model ranges and suppressed when inputs become stale.
• Clear member/adviser/facilitator turn ownership and deliberate keyboard handover, preserving canonical role, mandate and consent guards.
• Existing representative workers visibly attend paid handoff briefings; briefing holds routes without counting as rest or conferring plant authority.
• Honest customer targets and current-shift dispatch, with retained compensation and delivered recovery separated from current obligations.

Validation

• All four completed passes have focused regressions, production builds and native desktop/mobile day/dusk/night inspection.
• Latest pass: baseline61, final66 tests; typecheck, lint, formatting, depth/shaders, bundle, reachability and community checks pass.
• Exact latest build431 files,178119304 bytes(169.87MiB), below unchanged170MiB ceiling.
• Latest36 selected native Metal views, original readiness/quality guards, muted with no audio context. Actual public game controls, no actor/inventory overrides. Original camera failures retained, eight corrected camera-only views reviewed.
• Design detector retains one pseudo-selection background false positive, triaged against actual neutral-root computed styles.
• Four-commit secret scan clean; no outgoing blob exceeds100MB. No new dependencies, assets, saved schema or canonical role/consent/plant transitions.
• Authored and self-reviewed by the root. Synthetic role-play is not independent human participation. Pacing remains separately open; player/device/manual checks deferred.

Nell requested these completed changes live, followed by further core cooperative-learning refinement. Preserve normal CI, Pages release checks and archived versions. Protected dirty primary checkout untouched.