Changed -compactor.max-compaction-time default from 0s to 1h

[pracucci]

What this PR does

I propose to change -compactor.max-compaction-time default from 0s to 1h, to aim for a better fairness between tenants (when running a small number of compactor replicas) and address #2302 too.

I think enabling it (and setting to 1h) doesn't have any significant downside. For example, let's consider a single tenant cluster, with very long compactions. After 1h in a compaction cycle, the compactor stops compacting more blocks. However, since -compactor.compaction-interval=1h, the ticker already put a message in the channel so the next compaction cycle starts right after the previous one ended because the 1h limit has been reached. Basically, excluding the time it takes to scan the bucket, there's no expected "delay" introduced by setting -compactor.max-compaction-time when it triggers. Makes sense?

Which issue(s) this PR fixes or relates to

Part of #2302

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[pstibrany]

Given our default compaction interval is 1h, it would make sense to me to set max-compaction-time for tenant to the same value. But 2h is good start too.