Fix play-start jank and import hangs for material-heavy scenes

[papiguy]

Engine-side fixes surfaced by the TinySkies playground port — a material-heavy procedural globe game that exposed startup stalls, in-flight jank, and an import-time hang. None are game-specific; they benefit any large OSS scene.

• featureFlags: enable shared @import scripts in OSS unconditionally. REACT_APP_SCRIPTS_ENABLED is unset everywhere, so every behavior that @imports a helper failed to import. Scripts are a first-class OSS authoring feature; integrated keeps the env opt-in.
• Editor / RenderEvent / RenderingQualityModule: make the scene's serialized useShadows the source of truth and seed the renderer's shadow tracking from it. A no-shadow scene was paying shadow-variant shader compiles + a full-scene shadow pass every frame, and the first post-(re)create frame spuriously marked every material needsUpdate. Adaptive quality may still downgrade shadows but never force them on.
• EngineRuntime: precompile material pipelines with compileAsync before the loop starts, and hold the loading overlay until the first frame renders — folding one-time shader compiles into play-start load instead of streaming jank in mid-play. Both best-effort/bounded.
• GameManager: init behaviors in small batches that yield to paint, so the loading overlay advances instead of the UI freezing during a large world build.
• BehaviorManager: drop the per-behavior fixedUpdate "not implemented" warning — dozens of warn-with-stacktrace calls stalled play start for no benefit; not implementing an optional hook is normal.
• Editor.runScriptImport: pause the editor render loop during a bulk script import so per-command cost stays flat (was ~O(n^2), a multi-minute hang that could OOM the renderer).
• remote-go thumbnail: guard the /api/Scene/Edit POST behind !IS_OSS; OSS has no such endpoint, so it 404-spammed on every import.

Adds the port planning doc under docs/planning/.