Spark indexer fails with spark coordinates with "old" hadoops

[drcrallen]

When spark is built to be compatible with very old versions of hadoop, jets3t 0.7.x is set as the jets3t version. This version breaks horribly with many other things.

Setting the spark coordinates to something which has a more recent version of hadoop (2.x or higher) allows jets3t version 0.9.x to be brought in instead. These versions seem to work better with druid.