AbstractTrees should not be "opt out"

[ExpandingMan]

Ok, I have started digging into Keno's overhaul of this package.

So far the only major problem (at least according to me, we'll have to debate this) that I have noticed is that the fallback of children is the identity if the object is iterable.

One of the side effects of this was already causing problems: the package had to explicitly "opt out" for objects such as String which are iterable but not trees. That means the package has been "opt out" instead of "opt in" which seems pretty scary.

Anyway, I'm working through changing this now, I'll have more to say on what the implications of this are after I've gotten a bit further.

[ExpandingMan]

I'm waffling on this a bit.

I think it's fine for iterables to be trees, but only if it works such that non-iterable types are themselves detached roots without needing to define any special methods.

This may already be the case...

I'll report my conclusion on this when I make my PR, in the meantime, this is just a notice that this will be an issue of discussion.

[ExpandingMan]

My conclusion was ultimately that making every iterable a tree by default was a bad idea, and my PR eliminated the practice. Judging by the reactions to the PR, it seems we are mostly agreed on this, and the PR has since been merged, so I will close this issue.