use denser random graphs in distance tests

[sdmccabe]

See #239.

[leotrs]

at this point I'd recommend aliasing i.e. doing something like create_graph = nx.fast_gnp_random_graph(...) at the top of everything so we're sure we caught all of the instances of 0.1 and changed them to 0.3. I'm approving anyway...

[sdmccabe]

The tests are failing on denser graphs.

=================================== FAILURES ===================================
_____________________________ test_directed_input ______________________________
    def test_directed_input():
        with warnings.catch_warnings():
            warnings.filterwarnings(
                "ignore", message="Coercing directed graph to undirected."
            )
            G = nx.fast_gnp_random_graph(100, 0.3, directed=True)
    
            for label, obj in distance.__dict__.items():
                if isinstance(obj, type) and BaseDistance in obj.__bases__:
                    dist = obj().dist(G, G)
>                   assert np.isclose(dist, 0.0)
E                   assert False
E                    +  where False = <function isclose at 0x7f8a4b1d2268>(nan, 0.0)
E                    +    where <function isclose at 0x7f8a4b1d2268> = np.isclose
test_distance.py:88: AssertionError

[leotrs]

Do we know which is the offending distance?

[sdmccabe]

QuantumJSD is the offending distance.

[sdmccabe]

My knee-jerk reaction is that QuantumJSD should coerce to undirected, since the definitions in the paper I think tend to mention undirected graphs, and it's using the Laplacian.

This doesn't explain why upping the density caused a problem, of course.

[leotrs]

Anything that uses the Laplacian should coerce undirected tho.