Expose Dataset PartitionStrategy

It was mentioned on this discussion thread[1] that for operations on time partitioned datasets that require a specific granularity for operations that being able to determine the partitioning of a dataset would be useful.

So for example if you have an hourly based dataset and an Oozie job that ran on the 15 of every hour, you could know to zero out the minutes and seconds of the time for operations like removing partitions. Currently you have to rely on assumed code or naming conventions to know the specificity of partitioning a dataset.

[1] - https://groups.google.com/a/cloudera.org/d/msg/cdk-dev/UaQ9E63Cr0Q/GDqWsk3RwBoJ