Async support for the queries in the SQL Lab.

[bkyryliuk]

Added async support:

• Refactored the tasks.py to return the query id and added ability to run the code async.
• Changed the celery_tests to be more end to end (calling run_sql endpoint instead of celery function)
• Added implementation of async run of the hive and presto queries
• Added dependencies required by pyhive
• Implemented progress bar for the async queries
• Added extra fields to the Query object to improve traceability: limit_used, select_as_cta_used, executed_sql, select_sql

Tackles backend of the: #858, #746, #886

Implement progress bar for the presto / hive queries

Implement remote query execution

It is a preliminary PR and some more things needed to be done.
TODO:

• Unit test async queries (presto and hive engines)
• Implement cancel_query endpoint
• Add query_results endpoint

[clayallsopp]

Hi, does this still work? The SQL Lab queries are running on the celery worker, but (as the code is now) the results aren't returned and I'm not quite seeing how they get persisted anywhere. I tried forking and always returning the result from get_sql_results, but no dice either =\

[mistercrunch]

@clayallsopp async works for "CREATE TABLE AS" queries only, when enabled.

We're planning on adding support for result-backends probably next week, where the results of queries are persisted in a key-value store of your choice, probably S3 for us but we may support more backends out of the box. I have to figure out if flask-cache is the right abstraction for it, something else perhaps or we write our own...

The challenge is that the web request scope is short, and that we want to provide support for long-running queries even when don't have write access to the DB.