AggregatorFactory equals() and hashCode() methods.

Allow indexing tasks to specify extra classpaths.

Fix maxtime caching

This could be used by Hadoop tasks to reference configs for different clusters, assuming
that the possible configs have been pre-distributed to middle managers.

Update Druid api

Fixes DelimitedParseSpec serialization

explain what 'full cluster view' means

Fix extraneous timeseries zero filling at end of data interval

When the index's maxTime is not aligned with the query granularity, gran.next
can cause an time extra bucket to get zero-filled. Truncating first prevents
that.

query post-processing operators

CliHadoopIndexer: Modify maven arguments to be consistent with pull-deps.

fix granularity for precise periods with non-zero origin

- Allows post-processing of query results
- Example timewarp post-processing operator

HLL GroupBy fixes + tests