Allow cugraph-nx to run networkx tests for nx versions 3.0, 3.1, and 3.2

[eriknw]

This is one step to support #3770

Note that NetworkX version 3.2 is currently under development.

CC @rlratzel

[rlratzel]

LGTM, just had some ideas below.

[rlratzel]

(this is also a comment for all of on_start_tests(). No action for this PR beyond a potential FIXME if we like it.)

I'm assuming many backends will have potentially lengthy on_start_tests() implementations like this which use basically the same pattern and need frequent modification as the backends evolve to pass more tests.

I wonder if we could add a utility function to networkx which allows a backend developer to simply pass the incoming items and a YAML string and have them marked appropriately. The YAML could make it easy to define xfail messages for specific tests and nx versions, making these functions almost purely declarative and easier to read and maintain:

   from networkx.utils.backends import apply_xfails_from_yaml

   @staticmethod
   def on_start_tests(items):
      yaml = """
      nxver:
         3.2:
            test_betweenness_centrality.py:TestWeightedBetweennessCentrality.test_K5:
               weighted implementation not currently supported
      ...
      """
      apply_xfails_from_yaml(items, yaml)

The YAML could be inline like above, or in a separate file that's more discoverable.

[eriknw]

Yeah, that sounds nice; we could look into cleaning this up and making it easier for any backend to mark tests as xfail.

This seems pretty low priority for now--we'll see if I/we get frustrated by fiddling with this as we add more algorithms, and maybe we'll get motivated to make it better.

[rlratzel]

/merge