The default number of trials in rewire() should be proportional to the edge count

[szhorvat]

The default number of edge switching trials in rewire() should be proportional to the edge count. Currently this is fixed at 1000.

This is important, since we're trying to lead inexperienced users on the right track.

It led to a major misunderstanding in this recent preprint, see footnote * in Table 2. I just emailed the authors.

As a start I suggest using 10 times the edge count until we get arguments for a better factor than 10. The preprint I linked mentions Milo et al recommending 100, but I wouldn't put much weight on that: it's very old and ultimately it's a response to a criticism of their original biased randomization method, not actually focusing on edge switching methods.