Automatically set flume path from the partition strategy

Currently, sending records to partitioned kite datasets via flume requires the user to configure the HDFS file path by hand and match the partition strategy. From the examples:

tier1.sinks.sink-1.hdfs.path = /tmp/data/events/year=%{cdk.partition.year}/month=%{cdk.partition.month}/day=%{cdk.partition.day}/hour=%{cdk.partition.hour}/minute=%{cdk.partition.minute}

This leaves a lot up to the user and isn't clearly documented. Ideally, users would set a kite dataset sink, which handles partitioning.