Remove ivar documentation if possible

[jeromekelleher]

We use ivars in a few places in the documentation, but a nasty sphinx bug (sphinx-doc/sphinx#2549) means that they really don't work if you want to have nitpick mode on. We have a workaround monkey patched into setup.cfg.

If we move to JupyterBook, this won't work, as we don't have access to setup.cfg.

I've spent quite a bit of time on this just now in msprime, and the simplest thing to do is to use dataclasses instead and to document each variable separately, like the Population class here:

tskit-dev/msprime#1534

Ideally, we'd do the same here (although we might not have the freedom of using dataclasses)

[hyanwong]

Sphinx bug is fixed, and we've moved to JB. Can this be closed? We probably want to get around to remove some of the ivars anyway (see #1975) but there's presumably no urgency now?

[jeromekelleher]

Yes, let's close this.