Discovery of available query endpoints by clients -- LilyClient.getSolr()

[bruno]

Currently we expect clients to perform queries by directly contacting SOLR.

We might make this somewhat more convenient by publishing the location of SOLR in ZK and provided a method like LilyClient?.getSolr() to offer the possibility to talk to SOLR. This way information about SOLR does not need to be coded in or configured for every client application.

This is an alternative for #72.

Need to consider:

• that people might search on different SOLR nodes than those they index on
• that their might be multiple slave copies to use for querying
• that their might be multiple shards

So this seems like something an admin would set up, but still has the advantage that this info could then be used by all clients.