Add a method to make a graph from a complex matrix

[k-tamuraphys]

What is the expected enhancement?

Add a method like from_complex_adjacency_matrix , for makiing a graph from an adjacency matrix whose elements can be complex values.
I expect this method to work as follows:

adjacency_matrix = np.array([
    [1.0, 1.0+1.0j, -2.0j], 
    [0.0, 2.0, 1.0+1.0j], 
    [0.0, 0.0, 3.0]
])

graph = retworkx.PyGraph.from_complex_adjacency_matrix(adjacency_matrix)
list(graph.weighted_edge_list())

output:

[(0, 0, 1.0),
 (0, 1, (1+1j)),
 (0, 2, (-0-2j)),
 (1, 1, 2.0),
 (1, 2, (1+1j)),
 (2, 2, 3.0)]

When the input matrix is not symmetric, it may be convenient to make a graph from the elements of the upper triangular part of it, similar to from_adjacency_matrix.

[mtreinish]

So I have a draft locally implementing this but the only thing in your example which I will not able to do is reproduce the types in the output edge list. In your example for the real numbers they're floats in the output weight list, but when retworkx looks at the underlying numpy array it will see a 3x3 array complex128 and in the retworkx from_complex_adjacency_matrix() that complex number will be added as a complex so it will be (1.0+0j) not 1.0.

Other than that the PR is ready to go however unfortunately it's blocked on PyO3/pyo3#1798 and won't work yet. I will push it up as a WIP for the time being (and it'll fail on the msrv tests). So unfortunately I don't think this will make 0.10.0 as it's blocked on a pyo3 release.

[k-tamuraphys]

Thank you for the quick response. Having the output type (1.0+0j) would not matter to me.