Work around GEOS issue in voronoi_frames

[martinfleis]

While using the new voronoi_frames on a fairly large dataset, I noticed this issue libgeos/geos#1062 plus another where some polygons coming from GEOS were not actually valid, likely due to the same reason. This is a workaround that solves all I faced.

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 2 lines in your changes are missing coverage. Please review.

Project coverage is 84.9%. Comparing base (018f1e2) to head (72571b2).
Report is 7 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #693     +/-   ##
=======================================
+ Coverage   84.7%   84.9%   +0.2%     
=======================================
  Files        141     141             
  Lines      15200   15206      +6     
=======================================
+ Hits       12880   12913     +33     
+ Misses      2320    2293     -27     

Files	Coverage Δ
libpysal/cg/voronoi.py	72.5% <50.0%> (-0.7%)	⬇️

... and 6 files with indirect coverage changes

[jGaboardi]

Is this something that might be fixed in the future upstream? If so, should we make a ticket here to check the behavior in the future? Or will the workaround continue to be functional after the upstream issue is resolved?

[martinfleis]

I suppose it will be fixed in upstream. The if clause will be just skipped in that case, like it is in most cases where the issue does not happen right now. This is a very corner case but it was causing an issue when running enclosed_tessellation in momepy.