Closes #927, #928: Schema refresh improvements.

[emtwo]

This PR makes a variety of changes to fix bugs and more gracefully handle schema processing for data sources with many tables. A summary:

• Batch insert and update of column and table metadata
• Time limit on schema refresh increased to a soft limit of 5 min and hard limit of 10
• Column types are truncated to fit 255 character limit
• Fixed bug with str() function on TableMetadata accessing a non-existent variable
• Refresh schema every 12 hours instead of 30 min
• Check whether we should collect data samples before we spin off tasks to do so
• When a user requests a refresh, do it asynchronously
• Don’t force a schema refresh when there is no schema data to display (this clogs up the queue), just wait for the next refresh
• Changes to data sources (in data_sources.py) should trigger a refresh for the changed data source only (instead of all data sources)

[washort]

nice improvement, r+

[washort]

With the faster schema refresh do we still want this longer interval?

[emtwo]

It's not necessary at the moment, but I'm being cautious because the column examples are still a bit slow since they do select * from table limit 1 for each table. We can always turn it back up if it seems too slow I think.