Show only the CCTV cameras in view, thinned to one per ground cell when zoomed out

The CCTV layer kept every loaded camera live: each view change touched every
icon, every notification rebuilt every camera's state, and the picker listed
the whole catalog. With the default 4,000 cameras, a continental drag ran at
152-179 fps with 18-24 frames over 30 ms (worst 112 ms). It now runs at
235 fps with none. The layer now:

- shows icons only on or near the screen, and when zoomed out keeps one camera
  per world-anchored cell (healthy feeds first), so panning keeps the same
  cameras;
- grounds and draws coverage only for the cameras in view, the active camera
  and its coverage neighbours, retrying a camera at most twice per view regime;
- reruns the view pass while the camera moves, paced by its own cost, and
  projects with one matrix per pass instead of a Cesium call per camera;
- keeps the billboard buffer-usage hint fixed, so showing or hiding a few
  icons no longer rewrites the vertex data of every billboard;
- lists the cameras in view (active camera first) in the picker, closing with
  the count of cameras outside the view, and builds the full camera list in
  the UI state only when it is read.

src/layers/cctv/state.js

@@ -76,6 +76,63 @@ export function createState({ services }) {
 
   layerState._scratchFocusScreen = new Cesium.Cartesian2();
 
+  layerState._scratchViewScreen = new Cesium.Cartesian2();
+
+  layerState._scratchViewProjection = new Cesium.Matrix4();
+
+  // Counts view passes; a record shown by the latest pass carries its number.
+
+  layerState._viewPassId = 0;
+
+  // Records whose icons the last view pass showed (see refreshHorizonCulling):
+  // geometry refinement, the tiles-ready completion pass and ambient cards
+  // work over these, not the whole catalog.
+
+  layerState._viewRecords = [];
+
+  // Coverage ids refreshCoverageStyles last drew: part of the geometry
+  // working set even when their icons are thinned out.
+
+  layerState._coverageVisibleIds = new Set();
+
+  // View pass triggers besides camera settle: an interval while the camera
+  // moves, and a canvas resize observer.
+
+  layerState._viewMotionTimer = 0;
+
+  layerState._viewResizeObserver = null;
+
+  layerState._viewResizeCanvas = null;
+
+  // The camera the moveStart/moveEnd listeners are attached to.
+
+  layerState._viewListenerCamera = null;
+
+  // A motion pass changed the shown set and the settle pass has not yet told
+  // subscribers; the camera pose the last view pass ran at.
+
+  layerState._viewUnpublished = false;
+
+  layerState._viewPose = null;
+
+  // { at, ms } — when the last motion pass ran and what it cost.
+
+  layerState._viewMotionPass = null;
+
+  // Retries the view geometry a drain left unresolved, once cells warm.
+
+  layerState._viewGeometryRetryTimer = 0;
+
+  // { key, count } — the active camera's coverage-overlap count, which the
+  // summary reads on every notification.
+
+  layerState._neighborCountCache = null;
+
+  // { view, active, list } — the camera picker's index uiState publishes,
+  // rebuilt only when the view pass or the active camera changes.
+
+  layerState._cameraIndexCache = null;
+
   // Staggered geometry-load queue state (see startGeometryLoadQueue).
 
   layerState._geoQueue = [];