the haversine distance is no use for DBScan when Dimension <= 15000

[TanYufei]

method fit() in DBScan:

RadiusNeighbors rnModel = (new RadiusNeighbors(this.data, (new RadiusNeighborsParameters(this.eps)).setSeed(this.getSeed()).setMetric(this.getSeparabilityMetric()).setVerbose(false))).fit();
                this.info("fit RadiusNeighbors model in " + rnTimer.toString());

code for choosing BALL_TREE or KD_TREE to use:

private static BaseNeighborsModel.NeighborsAlgorithm delegateAlgorithm(RealMatrix arm) {
            int mn = arm.getColumnDimension() * arm.getRowDimension();
            return mn > 15000 ? BALL_TREE : KD_TREE;
        }

So, we can not use haversine distance when mn<=15000.(KD_TREE do not suport haversine distance)