Improvements towards building a consistent spellcheck index

[cbeer]

In production, we trigger building the spellcheck index with a cronjob that runs nightly. This job takes ~1 hour to complete. There are several problems with the current approach:

• [ ] the HTTP request to build the index times out before the index is complete (although the index appears to usually be successfully built)
• [ ] the HTTP request is sent to the load-balancer and redirected to any solr node hosting the collection. Unfortunately, the spellcheck index is not replicated across the cluster, so the command has to be run on every node that hosts the collection.

[mejackreed]

Trying to break out a class that will do the Solr Collections API stuff to accomplish the second bullet point.