Add prometheus metrics about the latency on the task queue

[irevoire]

It should be done after #5044; otherwise, there will be a ton of conflicts.
It should also probably be done using the work of this PR: #5048

We should add the duration between now (at the time of the polling) and the enqueuedAt time of the oldest enqueued or processing task.
That gives us a measure of the global indexing latency of the engine.
To retrieve the oldest enqueued task, we can do a filtered query to the task queue with a limit of 1 and the reverse parameter introduced in the PR above.

[takaebato]

I'll work on this!
→I'll submit the PR by Saturday!