1) Initial commit of refactorings on top of housejester's to simplify the zipping and simplify the creation of other methods of loading data

1) Move the local cacheFile logic out of the S3 pullers into the SingleSegmentLoader
2) Make the S3SegmentPuller just pull down the file
3) Make the Loader do the unzip, ungzip, or rename
4) 2 and 3 make S3ZippedSegmentPuller not necessary (still there, just deprecated and empty)
4) Tweak the TaskToolbox so that the Pullers returned by getSegmentGetters behave the same as they did before

1) Fix alert sent from line DruidMasterBalancer:71 to not include all of the segments that a server is holding