Remove OneAnd?

OneAnd doesn't seem very useful to me now that Cats has first-class NonemptyList and NonemptyVector types. It can be awkward to work with, especially if you don't have type aliases and "companion" object set up for the type that you are working with.

It keeps its head element separate like a cons-list, making it not very efficient to use with a vector-like structure, but it combines by delegating to an Alternative instance, which isn't efficient for list-like structures.

People might reach for OneAnd when they want a non-empty version of a structure that doesn't have first-class non-empty support in Cats. Some common types that they might reach for are non-empty sets and non-empty maps. However, the OneAnd model doesn't actually work for those (see #1909 for an example).

I'm not sure that OneAnd cuts it as an abstraction, and I don't know anyone who is actually using it. I vote for removing it. Is there anyone who uses it and/or has reason for it to stay?

Here is a related gitter discussion.

Start by reviewing the OneAnd implementation and its usages, then read the related discussion in issue #1909 and the linked Gitter conversation. The issue does not name specific files or tests; completion would require an agreed decision on whether OneAnd should be retained or removed and a defined migration scope.