Allow user to select and use existing repo rather than create new one for each data set

[olivierthereaux]

Summary

Octopub creates a repository for each data set. Give user the option to publish data set to an existing repo so as to keep number of repo in the user's account in check.

Context

Currently, each time a user uploads and publishes a new data set, a new repository is created for the data set, complete with Jekyll files. This is somewhat undesirable as it means that a user publishing a number of sets would end up with a lot of repository. While there are no hard limits on public repositories on GH, it makes searching and management of one's repos cumbersome.

My understanding of Octopub archeology tells me that this decision went as follows:

• The repo has to be writable by the user (as octopub publishes on their behalf) and publishing to a repo owned by the user makes sense. Also, legal/license considerations.
• It is probably easier to create a blank repo and write all files to it than deal with potential conflicts when writing to an existing repo (name clash, different version of Jekyll files)
• #204 was certainly a factor - if some datasets are going to require a private repo and others a public repo, why not simply keep each set in a different repo.

Why change

• Because #204 is essentially mooted by the new paradigm of pre-publish/publish. The ability to write to a private repo is on its way out.
• Because it would offer the user a welcome additional degree of control of how the data collections are organised.

How to do it

TBD. I can envision:

• a "use existing repository" option control at the end of the publishing flow
• ... which may require some checking of which repos are already octopub'd, so as not to suggest writing into a non-octopub repo?
• need to check whether API allows repository listing/traversal. Most likely so.