BUG: misaligned weights in the gabriel graph

[martinfleis]

Found a bug in Gabriel graph. heads and tails come unsorted from filtering, resulting in misaligned weights. See repr:

import geopandas
from libpysal.graph import Graph
import momepy

blgs = geopandas.read_file(momepy.datasets.get_path('bubenec'), layer='buildings')

delaunay = Graph.build_triangulation(blgs.centroid, "delaunay", kernel="identity")
gabriel = Graph.build_triangulation(blgs.centroid, "gabriel", kernel="identity")
relative = Graph.build_triangulation(blgs.centroid, "relative_neighborhood", kernel="identity")

>>> delaunay[48]
neighbor
46      24.357863
47      32.418353
49      49.764015
74     155.698588
75      99.914953
141     93.302955
Name: weight, dtype: float64

>>> gabriel[48]
neighbor
46    128.634198
47     24.117386
Name: weight, dtype: float64

>>> relative[48]
neighbor
46    24.357863
47    32.418353
Name: weight, dtype: float64

It should be a subset of Delaunay, not alter the value.

[ljwolf]

This looks okay to me!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.9%. Comparing base (bcabdbc) to head (de59f61).
Report is 12 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #709     +/-   ##
=======================================
- Coverage   85.0%   84.9%   -0.1%     
=======================================
  Files        141     141             
  Lines      15203   15232     +29     
=======================================
+ Hits       12924   12926      +2     
- Misses      2279    2306     +27     

Files	Coverage Δ
libpysal/graph/_triangulation.py	98.0% <100.0%> (+<0.1%)	⬆️
libpysal/graph/tests/test_builders.py	100.0% <100.0%> (ø)

... and 3 files with indirect coverage changes