Druid run_query executes phase 1 query with no limit

[michellethomas]

In the druid run_query here a Phase 1 query is created if timeseries_limit == 0, but that can result in an incorrect query if the limit field is not present and there are parts of the Phase 2 query that are necessary (but don't get added).

This causes problems if a viz type does not have a limit but does require a phase 2 query or you remove the limit on a line chart.

@mistercrunch would you suggest having different logic for determining phase = 1 or pulling the necessary logic out of phase 2 and into phase 1?

[michellethomas]

@mistercrunch any thoughts on this? If there's no timeseries limit we run a phase 1 query, which isn't always correct. Do you know why this step is necessary?

We are seeing group by line charts that should be 2 phase queries but the query only has a phase 1.

https://github.com/apache/incubator-superset/blob/master/superset/connectors/druid/models.py#L1072