hdfs-sunset

change not to fail with IOException

add --parallel option to control number of partitions

try to distribute huge files among executors

change to use dfs.blocksize as a read buffer size

initial implementation