Establish repo_utils subpackage

[mih]

A design goal of datalad-next is to shed the dependency on the ancient Repo interface in datalad-core. It was established when DataLad used a more stateful representation of a Dataset/Repo, and when Git and git-annex where less powerful. Today it is largely an obstacle.

Nevertheless, there continues to be a need to have implementations for common query and manipulation tasks for repositories. We should establish a place for such implementation to promote their reuse, over reimplementations of near-copies.

#592 brings a whole range of internal helper that should be promoted to the public, and equipped with tests.

It would be worthwhile to not require a standard data structure (like a Dataset) for such helpers, but work with the bare necessities (often just a path) to promote wide applicability.

Moreover, it would be helpful to aim for generic implementations that work with git and git-annex repos alike, and also in adjusted mode repos -- unless there is a speed difference that justifies alternative implementations, or if there is a substantial complexity cost to pay.