Double check and document (if needed) `.name` conventions for `iter_collections`

[mih]

I cannot put my finger on it, but I feel the nature of the value of .name for a reported item is non-uniform.

I am not aware of a documentation for the convention, if how iterators should behave.

The .name should always be a PurePath|Path objects.

However, should it carry the base path (key argument to the iterator) in it? Or not?

It would make sense to me to never have it. Rational:

There is no reason while a (file) collection should be on a file system. It could have a very non-Path base address and no hierarchical organization inside (think object store: some URL, plus some object ID).

We need to go through and verify that we did not overfit to a filesystem use case in an implementation.

[mih]

#581 brought up a related issue. It makes sense to have the name attribute be as cheap as possible, and only provide access to more complex/expensive types via lazily evaluated properties.