propagate relabelling for gabriel weights

[ljwolf]

Hey,

This solves the test bench for the Delaunay-based graphs, in the sense that ids are retained and keept aligned with the input dataframe. Everything prints the same:

import geopandas
from libpysal import weights, examples
examples.load_example("south.shp") # I can't locate south if I don't load it first, even if it's downloaded?
south = geopandas.read_file(examples.get_path("south.shp"))
south = south.set_geometry(south.centroid)

print(south.FIPS.iloc[:5].tolist())
for ids_ in ('FIPS', south.FIPS):
    wdel = weights.Delaunay.from_dataframe(south, ids=ids_)
    print(wdel.id_order[:5])
    wgab = weights.Gabriel.from_dataframe(south, ids=ids_)
    print(wgab.id_order[:5])
    wrn = weights.Relative_Neighborhood.from_dataframe(south, ids=ids_)
    print(wrn.id_order[:5])

this is what I mean about "sorting by default." I don't think that, if the user provides ids='FIPS', we should silently return a mis-aligned weights object from from_dataframe(). We can do this independently of the parent W class.