Added new Cube I/O caching algorithm BoxcarCachingAlgorihthm designed...

Added new Cube I/O caching algorithm BoxcarCachingAlgorihthm designed specifically for the ProcessByBoxcar read/write pattern used by applications like gradient, in reference to issue #167.  This does not fix the issue entirely, as the performance gains do not bring the running time on test machines dwon from 3.5 minutes to 0.5, but do bring it down to around 2 minutes.  We will continue to look into ways to improve the performance of Cube, but if problems with running gradient on the cluster were predominately I/O related, then this should mostly resolve the issue.  If it turns out performance was bottlenecked by CPU, however, then this fix will not resolve the issue.

git-svn-id: http://subversion.wr.usgs.gov/repos/prog/isis3/trunk@2801 41f8697f-d340-4b68-9986-7bafba869bb8