Exponential backoff for query retry has no upper limit

[mreid-moz]

Issue Summary

Per Bug 1565659, a scheduled query failed several times in a row and got itself into a state where, due to exponential backoff, it would not try again for 6 months. For a query that is expected to run every day, this is no good.

An upper bound of 24hrs (or the query refresh interval, whichever is larger) seems reasonable.

Steps to Reproduce

1. Write a query, run it on a scheudule
2. Cause this query to fail several times in a row (exercise to the reader)
3. Fix the thing that was causing failures
4. wait 6 months for the next scheduled update.

Technical details:

• Redash Version: prod deployed at sql.telemetry.mozilla.org
• Browser/OS: Nightly 70 / MacOS Mojave
• How did you install Redash:

[jezdez]

This would definitely be a feature we would want to propose upstream since it's a pretty central part of Redash and I'd rather not mock around with those internals in the fork.