Active Fedora

This is a mix of improvements to ActiveFedora and improvements to Shelver, all centered around allowing for efficient, reusable, configurable solr indexing.

Primary motivation: this will make shelver much more efficient when indexing objects that conform to multiple models.

• Rely on class level methods (and nokogiri) to parse xml directly from Fedora datastreams into Solr documents
• * Model#to_solr( pid ) relies on the XML in Fedora (bypassing the step of loading an instance of the Model)
• * instance.to_solr could either rely on the class method (which means you should always save before calling to_solr) or could rely on internal hashes
• Adjust implementation of NokogiriDatastream accordingly to support this
• Re-work shelver to take advantage of these improvements

this also relates to the work on creating sane configs (#706)