Adding floating point considerations to tutorial

[amcandio]

Floating Point Considerations

As discussed in previous issues and PRs (e.g., #4972, #4592, #8316), users sometimes face unexpected results when using floating point values such as edge weights or capacities. These issues are often caused by rounding errors, not bugs in the algorithms.

The moment floating point numbers are used, all results become approximate. So to avoid confusion, we update tutorial to clarify these considerations.

[rossbar]

Thanks @amcandio , I think this is a step in the right direction. One idea that occurred to me when considering how to better document numerical precision was to add a standard reference document (similar to what is done for seeding/randomness) and then link to this reference doc from the functions where precision issues are prominent.

That's too much for this PR though - I'm +1 for getting this information somewhere visible in the docs and this seems a sensible place to me!

One style nit - could we get the line limits down to something more reasonable (80 or 88 or 110 chars). In principle the linter should be doing this file is clearly not being picked up 🙃

[amcandio]

Thanks! I fixed the line length. We can add a more-detail doc like https://networkx.org/documentation/stable/reference/randomness.html in a separate PR.

[dschult]

Thanks @amcandio

[rossbar]

Thanks @amcandio !