Abstraction layer for performing queries (no direct SOLR use)

[bruno]

Currently we expect clients to directly perform queries on SOLR.

There can be good reasons to put some layer in front of this:

• we might want to do result set building based on data fetched from Lily rather than from SOLR. This would remove the need to push certain fields to SOLR or to store them in non-indexed form in SOLR. This is even more interesting when using link dereferencing. It might also remove the need for SOLR's document cache (not sure, it might also be used for other purposes like faceting).

• translate from logical names to IDs, e.g. to search in a certain vtag view (see #58)

• adding access control (by e.g. automatically enforcing some query conditions)

Note that most of these thing would probably also be possible by adding plugins to SOLR, and might in some cases make things possible which are difficult to do afterwards.