GitLab should only block unpausing Elasticsearch indexing if there is migration running that requires indexing to be paused

Problem

Noticed in https://gitlab.zendesk.com/agent/tickets/208089 that a user paused indexing and then could never unpause it because their migrations weren't running. Their migrations weren't running because their index was created before %13.0 and to resolve the issue they need to recreate the index from scratch.

Their problem can be solved by recreating the index and they should do that anyway but there are other cases where our logic is not quite right. The only reason we block unpausing indexing is that certain migrations require the indexing to be paused. Some migrations do not require indexing to be paused. And those migrations may take a very long time to run. If a user pauses indexing during a long running migration that does not require indexing to be paused then they will be unnecessarily prevented from unpausing indexing. Even worse, they may be permanently blocked from unpausing indexing as certain migrations like !59836 (merged) will require indexing to be unpaused to make progress (live lock).

Solution

Update the code which blocks unpausing to actually check that the currently running migration requires indexing to be paused.