Currently, we always create a new cluster object even if the existing cluster origin had no neighbors to merge with on lower zoom level. This is redundant for memory footprint, and also makes it impossible to identify the same cluster across multiple zooms by the same cluster_id. This PR fixes this by carrying cluster objects (that don't receive new points) onto lower zooms. Closes #189.
Currently, we always create a new cluster object even if the existing cluster origin had no neighbors to merge with on lower zoom level. This is redundant for memory footprint, and also makes it impossible to identify the same cluster across multiple zooms by the same
cluster_id. This PR fixes this by carrying cluster objects (that don't receive new points) onto lower zooms. Closes #189.