fix: resolve 10 defects uncovered by Android Maps Testing Toolkit

[dkhawk]

Summary

Fixes 10 production bugs across :data, :heatmaps, :clustering, :library, and :ui uncovered during integration and adversarial verification with the Android Maps Testing Toolkit (v1.1.0-rc01), forked from feat/migrate-ktx-to-utils (#1716):

#	Module	Issue Fixed
1	:data (KmlLayer)	Cache KmlGroundOverlay -> ModelFeature in mGroundOverlayMap so removeLayerFromMap() removes ground overlays from MapViewRenderer's IdentityHashMap instead of leaking them on GoogleMap.
2	:data (GeoJsonLayer, KmlLayer, MapViewRenderer, Style)	Propagate clickable (default true) and visible to PolylineOptions and PolygonOptions so GeoJSON and KML polygons and polylines are clickable and dispatch OnFeatureClickListener callbacks.
3	:data (GeoJsonLayer, KmlLayer, MapViewRenderer)	Forward markerManager, polygonManager, polylineManager, and groundOverlayManager to MapViewRenderer collections and attach click listeners to those collections so layers coexist with ClusterManager / MarkerManager without clobbering global GoogleMap listeners.
4	:data (KmlLayer, GeoJsonLayer, MapViewRenderer, Style)	Aggregate placemarks and ground overlays from <Document> containers in hasPlacemarks(), getPlacemarks(), features, and getGroundOverlays(), and preserve styles on <MultiGeometry> / GeoJsonMultiLineString / GeoJsonMultiPoint / GeoJsonGeometryCollection via CompositeStyle.
5	:data (GeoJsonLayer, MapViewRenderer, Style)	Propagate GeoJsonPointStyle title, snippet, isDraggable, isFlat, isVisible, infoWindowAnchorU/V, and custom icon (BitmapDescriptor) to MarkerOptions and AdvancedMarkerOptions.
6	:heatmaps (HeatmapTileProvider)	Clamp zoom when indexing maxIntensity (preventing ArrayIndexOutOfBoundsException at zoom 22) and clamp bucketX/bucketY to 0 until gridDim so points on the upper inclusive boundary maxX/maxY do not throw ArrayIndexOutOfBoundsException.
7	:heatmaps (HeatmapTileProvider)	Check && wrappedPoints.isEmpty() before returning TileProvider.NO_TILE so cross-antimeridian points render across the International Date Line, and enforce MIN_RADIUS..MAX_RADIUS (10..50) and 0.0..1.0 validation in setRadius and setOpacity.
8	:clustering (DefaultClusterRenderer)	Clear stale marker.title and marker.snippet when a ClusterItem's title/snippet is updated to null, and update marker.zIndex in onClusterItemUpdated even when position is unchanged.
9	:library (MapObjectManager)	Invoke setListenersOnUiThread() synchronously when constructed on the main thread instead of unconditionally posting to the back of the main looper queue.
10	:ui (AnimationUtil, IconGenerator, RotationLayout)	Snap directly to finalPosition when durationInMs <= 0L (preventing 0 / 0.0f = NaN), clamp t to [0f, 1f] so AccelerateDecelerateInterpolator never rebounds backward on frame overshoot, and use degrees.mod(360) so negative multi-turn rotations (<= -450°) do not throw IllegalStateException.

Testing

• Added unit regression tests across MapObjectManagerTest, ClusterManagerTest, KmlLayerOnMapTest, GeoJsonLayerOnMapTest, HeatmapTileProviderTest, AnimationUtilTest, and IconGeneratorTest.
• Verified all unit tests pass across :library, :clustering, :data, :heatmaps, :ui, and :lint-checks, as well as the full 5-module Android Maps Testing Toolkit (v1.1.0-rc01) test suite.

[googlemaps-bot]

Code Coverage

Overall Project	56.94% -2.4%	🍏
Files changed	62.71%	🍏

Module	Coverage
Kover Gradle Plugin XML report for :library	89.59% -0.15%	❌
Kover Gradle Plugin XML report for :heatmaps	88.62% -0.19%	🍏
Kover Gradle Plugin XML report for :ui	80.78%	🍏
Kover Gradle Plugin XML report for :data	54.29% -3.84%	🍏
Kover Gradle Plugin XML report for :clustering	33.26% -1.51%	❌

Files

Module	File	Coverage
Kover Gradle Plugin XML report for :library	MapObjectManager.kt	94.05% -3.97%	❌
Kover Gradle Plugin XML report for :heatmaps	HeatmapTileProvider.kt	92.61% -0.27%	🍏
Kover Gradle Plugin XML report for :ui	AnimationUtil.kt	92.7%	🍏
	IconGenerator.kt	90.34%	🍏
	RotationLayout.kt	88.97%	🍏
Kover Gradle Plugin XML report for :data	Style.kt	100%	🍏
	MapViewRenderer.kt	76.13% -16.55%	🍏
	GeoJsonLayer.kt	55.37% -10.74%	🍏
	KmlLayer.kt	51.35% -17.09%	❌
Kover Gradle Plugin XML report for :clustering	DefaultClusterRenderer.kt	17.25% -1.34%	❌
	ClusterRendererMultipleItems.kt	0.15% -2.33%	❌
	DefaultAdvancedMarkersClusterRenderer.kt	0% -2.92%	❌

[kikoso]

What happens here if someone calls setMap() with a different map? The managers we stored in the constructor are still bound to the old GoogleMap, so it looks like the features (and the click listeners) would ended up in the old map. Before this PR setMap was ignoring the managers, so this case used to work. What if we drop the managers (or throw) when the map is not the same one?

[dkhawk]

Good catch! Updated setMap() to remove existing features from the previous renderer first and drop the constructor-supplied managers/collections when map changes to a different GoogleMap instance.

[kikoso]

Same question than in GeoJsonLayer.setMap(), what about the managers here when the map changes?

[dkhawk]

Updated KmlLayer.setMap() with the same behavior to clean up the previous renderer and drop the managers/collections when switching to a different GoogleMap.

[kikoso]

Every time the renderer is recreated (e.g. setMap(null) + setMap(map)) we call newCollection() again, and there is no way to remove a collection from the manager. Could we reuse the collection instead of creating a new one each time?

[dkhawk]

Good point! Lazily created the manager Collection instances once per layer and passed them into MapViewRenderer so setMap(null) + setMap(map) reuses the existing collections instead of calling newCollection() again.

[kikoso]

This changes what getPlacemarks() returns: before it was only the top-level placemarks, now it recurses into all the containers. What about apps that already walk getContainers() themselves (like KmlDemoActivity)? They would get the nested placemarks twice. What if we keep the old behavior here and add a new getAllPlacemarks() with KDoc? Same would apply for hasPlacemarks(), features and getGroundOverlays().

[dkhawk]

Agreed! Restored hasPlacemarks(), getPlacemarks(), features, and getGroundOverlays() to return only top-level items, and exposed getAllPlacemarks() and getAllGroundOverlays() with KDoc for recursive traversal across nested <Document> and <Folder> containers.

[kikoso]

PointStyle, LineStyle and PolygonStyle are public since v5.0.0, so adding constructor params is a binary break for the constructors and copy(). Fine for 6.0 I think, but should we add it to MIGRATION.md? The fix: title won't flag it otherwise.

[dkhawk]

Good call — documented the expanded PointStyle, LineStyle, and PolygonStyle constructor signatures in MIGRATION.md under 6.0.0 changes.

[kikoso]

Is it intended to have a BitmapDescriptor here? GeoJsonMapper describes this model as platform-agnostic, and this couples it to the Maps SDK.

[dkhawk]

Good catch! Removed BitmapDescriptor from PointStyle so com.google.maps.android.data.renderer.model stays completely platform-agnostic, and handled custom GeoJsonPointStyle BitmapDescriptors via an internal descriptor cache in MapViewRenderer.

[kikoso]

Adding a new subtype to the sealed Style will break any exhaustive when over Style in consumer code, could we mention it in MIGRATION.md too? Also, what about #1800 (explicitApi)? CompositeStyle, the new constructor properties and MapViewRenderer don't have public, so whichever lands second will need to fix it and regenerate data.api.

[dkhawk]

Added CompositeStyle (and its impact on exhaustive when expressions over Style) to MIGRATION.md, added explicit public modifiers here, and will rebase #1800 and regenerate data.api right after.

[kikoso]

What about subclasses that set their own title in onBeforeClusterItemRendered? When the item has no title and no snippet, this now sets marker.title = null and wipes it on the first update, before it was left untouched. Also, ClusterRendererMultipleItems and DefaultAdvancedMarkersClusterRenderer still have the old logic, should we update them too so the three renderers behaves the same?

[dkhawk]

Great catch! Updated onClusterItemUpdated so that when both item.title and item.snippet are null, marker.title and marker.snippet are left untouched (preserving custom titles set by subclasses in onBeforeClusterItemRendered), while clearing marker.snippet when only title or snippet is non-null. Also applied the same title/snippet and zIndex update logic to ClusterRendererMultipleItems and DefaultAdvancedMarkersClusterRenderer.