Add `keys` for `zip`

[bkamins]

The iterator created by zip does not have keys defined. I think it would be good to add it.

Without it e.g. paris on zip fails. Then in consequence all infrastructure using pairs like findall fails on the result of zip.

Is there any reason why this was not implemented? I could try propose a PR for this if we agree that it is good to have keys for zip.

[ararslan]

Zip iterators aren't indexable, so what are their keys? I'm not sure it's well-defined in this case.

[bkamins]

This is right, but the idea was that often you zip collections that support indexing. Then keys for zip would return tuples of keys of underlying collections. If at least one of zipped collections does not support indexing then keys on zip would error.

But I was not 100% sure if this is a good concept - that is why I have asked.

[rfourquet]

In #22489 is a related proposal to allow indexing into zip iterators.

[JeffBezanson]

Some might expect keys(zip(a,b)) == a, so I'd hesitate to do this.

See also #27855.

[bkamins]

I am OK to close it if it is not a clear cut thing (and it seems so).