Spread shared runner minute reset to 8h

What does this MR do?

Over the previous couple of months we experimented with this spread and 8hrs seems to be the best so far.

Spreading this over a period of time is required to prevent severe database service degradation. We tried 3hrs which didn't help much, we also tried 24h which wasn't any better than 8hrs.

Originally written by @cmiskell

1. November 2020: Original issue, no spread, it was really bad.
2. December 2020: 3 hour spread, things still melted
3. January 2021: 8 hour window improved things somewhat (per comment at gitlab-com/gl-infra/production#3268 (comment 477088591)), although still with impact
4. February 2021: 24 hour window: not really better than 8 hours initially, but once we could run the analyze regularly we kept it under control.
5. March 2021: 3 hour window, everything melted again.

Context #323069 (comment 522474466)

Related #323069 (closed)

Does this MR meet the acceptance criteria?

Conformity

• 📋 Does this MR need a changelog?
  • I have included a changelog entry.
• N/A Documentation (if required)
• Code review guidelines
• Merge request performance guidelines
• Style guides
• Database guides
• Separation of EE specific content