Large Collection Reindexing Fills Memory allocation

[davidmcclure]

The current implementation of SolrSearch attempts to create all objects before creating HTTP requests to Solr. This needs to be refactored to more efficiently iterate results without filling memory allocation (or requiring a very large memory allocation).

[erochest]

Evaluate this for inclusion in milestone 1. If Omeka can already handle the paging, that's fine; otherwise, as important as this issue is, we probably need to hold off until a 1.1 release.