Add numpy version

[vgoliber]

Update unittest for new test that is inconsistent.

Note: the unittest fails when num_new_cs = 1, so this has been updated to choose a number between 2 and 4 instead of 1 and 4.

[hhtong]

Do you know why the error is occurring when num_new_cs==1? I see that it's not that the test fails when num_new_cs==1, it's that an exception is raised in demo_np.py:

Traceback (most recent call last):
  File "C:\Users\circleci\repo\tests\test_integration.py", line 119, in test_same_bqm
    bqm_np = demo_numpy.build_bqm(potential_new_cs_nodes, num_poi, pois, num_cs, charging_stations, num_new_cs)
  File "C:\Users\circleci\repo\demo_numpy.py", line 86, in build_bqm
    q2 = quad_col[dist_mat != 0]
IndexError: boolean index did not match indexed array along dimension 0; dimension is 90601 but corresponding boolean dimension is 903

So it seems like something needs to be adjusted in the demo_numpy.py formulation, if num_new_cs==1 is a valid argument.

[hhtong]

If we do end up needing to enforce that num_new_cs>1, it should be explained in the docstring why.

And then if that's the case, in this test, I think we'd only need to do:

Suggested change

	num_poi, num_cs, num_new_cs = (random.randint(2,4), random.randint(2,4), random.randint(2,4))
	num_poi, num_cs, num_new_cs = (random.randint(1,4), random.randint(1,4), random.randint(2,4))

[hhtong]

Also, I see that in demo_numpy.py, when num_new_cs>1, the dist_mat calculated for constraint 2 seems to be ignored/overwritten by constraint 3. I'm not familiar with this formulation, so I wanted to double check with you that the new dist_mat equation for constraint 3 also takes constraint 2 into consideration. (My assumption is that it does, since the unit test you wrote passes in this case)

[vgoliber]

The distance matrix getting written over is correct because it's a different distance matrix each time - I just reused the same object.

[vgoliber]

@hhtong this should be resolved now.