Update multigraph docstrings to reflect remove_edges_from behavior.

[rossbar]

Addresses the first bullet of issue #5158 , i.e. updating the documentation to correct the stated behavior of remove_edges_from when 2-tuples are passed. The remaining parts of the issue are larger discussions that should probably be moved to their own topic.

[dschult]

This has clear implications for the doc_string of remove_edge as well. This remove_edges_from doc_string states clearly that it is the most recently added multiedge that is removed. While the remove_edge doc_string says in more than one place that it is an "arbitrary" edge that is removed. I think the difference is due to dict's now being ordered. But since we're improving this doc_string is it in scope to change it for the remove_edge doc_string as well?

Also, we should check the MultiDiGraph doc_string for remove_edges_from to make sure it is updated to match this doc_string. In this case, MultiDiGraph inherits this method. So the doc_string is identical. But that class' remove_edge doc_string will need to be updated from "arbitrary" to "most recently added".

[rossbar]

But since we're improving this doc_string is it in scope to change it for the remove_edge doc_string as well?

Agreed - this should be taken care of in 1efc66d and 39a93a3. As noted, the docstring for remove_edges_from is inherited, but the docstring for remove_edges is not, so I've tried to be careful and make the appropriate changes in both MultiGraph.remove_edges and MultiDiGraph.remove_edges.

Also, we should check the MultiDiGraph doc_string for remove_edges_from to make sure it is updated to match this doc_string.

As a quick sanity check, here's the html page for the inherited docstring. AFAICT this seems to be working as expected.

[dschult]

Looks Great! Thanks for this!