[API] Prepare "repository index" and incorporate it in API

[bzz]

Right now, in order to get URLs or the repositories, or languages we need to traverse all .siva files and extract original repo urls from config.

This issue is about speeding it up by pre-computing a dataset of original repo url, init_ hash, languages

It has 2 parts:

• prepare a dataset, using Spark API
• incorporate it into the Spark API so it is used for filtering, instead of traversing .siva files

[ajnavarro]

Related to https://github.com/src-d/engine/issues/166 and https://github.com/src-d/backlog/issues/795

[ajnavarro]

Superseded by src-d/sourced-ce#242