[SPARK-18590][SPARKR] Change the R source build to Hadoop 2.6

This PR changes the SparkR source release tarball to be built using the Hadoop 2.6 profile. Previously it was using the without hadoop profile which leads to an error as discussed in https://github.com/apache/spark/pull/16014#issuecomment-265843991

Author: Shivaram Venkataraman <shivaram@cs.berkeley.edu>

Closes #16218 from shivaram/fix-sparkr-release-build.