add achilles_processing table

[fdefalco]

A potential change in the way we manage achilles processing, this would add a table in which the execution details for each analysis are stored. This would include the analysis_id, execution_duration, count_rows_generated, execution_timestamp, error.

This could be leveraged to ensure only successful analyses are merged into the final table, whereas currently the final merge will fail because an analysis that fails does not create its temp table for the final merge.

[alondhe]

Another benefit will be not having to worry about stratum casts in WebAPI drilldown queries, which are currently failing in Redshift for us (and confirmed by @clairblacketer in your env).