[Numpy-discussion] Merging very limited weights support for quantiles/percentiles