Incompatible scipy version

[JoelPasvolsky]

Description If you pip install dwavebinarycsp and then build penaltymodel-maxgap, running any dwavebinarycsp.stich() fails for scipy==1.4.1:

decorator==4.4.2
dimod==0.9.11
dwave-networkx==0.8.4
dwavebinarycsp==0.1.2
homebase==1.0.1
networkx==2.5.1
numpy==1.20.2
penaltymodel==0.16.4
penaltymodel-cache==0.4.3
penaltymodel-lp==0.1.4
penaltymodel-maxgap==0.5.4
PySMT==0.8.0
scipy==1.4.1
six==1.16.0

For example:

>>> import dwavebinarycsp as dbc
>>> csp = dbc.factories.random_2in4sat(8, 4)
>>> dbc.stitch(csp)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>

<see referenced issue for more>

Upgrading scipy solves the problem

To Reproduce As described above

Expected behavior No errors

Environment:

• OS: Ubuntu 16.0 and Win10 and colab
• Python version: 3.7.0

Additional context See https://github.com/dwave-examples/factoring-notebook/issues/16 I was unable to reproduce using pip install instead of build penaltymodel from source. I noticed that colab builds from source and managed to reproduce the problem in both unix and Window OSs.