It gets the pointer's position in pixels from the top left of the map, as MapLibre's
event.point has it, and returns the objects there, the topmost first, or none.
It runs on every pointer move and click while a handler listens to the layer, so keep it
quick: project your objects' positions through the camera matrix you drew them with, and
compare in pixels.
It owns the depth. The SDK knows nothing of what stands in front of your objects, so a
screen projection also hits an object behind a 3D building.
Its hits rank against the map's other features by the layer stack: a handler on the 3D buildings,
which sit above a 'lowest3D' layer, takes the click on a building even where your object
stands in front of it.
Tells which of a custom layer's objects lie under a point on the map.