Add optimized cube materialization job

[shangyian]

This will aim to consolidate the different types of cube materialization into a single job that chooses the most efficient way to materialize the cube for downstream use. This will be a new materialization job, to avoid conflict with existing jobs. The job will take advantage of the ability to build queries that compute pre-aggregated measures for a set of metrics and dimensions from #1242

The core DJ API will need to build the following pieces of materialization job metadata:

• A list of measures queries: Each query computes the pre-aggregated measures for a specific subset of metrics and dimensions. For each measures query, we will additionally keep track of:
  • The node it was generated for
  • A list of measures and dimensions that it provides
  • Spark configuration (it is unclear how we would configure this at the moment)
• A combiner query: This query merges the results of the above measures queries into a single dataset.
• Druid ingestion spec: Druid-specific configuration for the combined dataset
• The temporal partition: This will need to be a shared dimension for all metrics in the cube.