Monitoring: Capture Task Management output

[elasticmachine]

Original comment by @pickypg:

Now that the task management API exists, we should monitor it to try to provide useful information to users.

For example:

• Commonly running tasks.
• Long running tasks.

It's important to remember that we poll every ~10 seconds (configurable, but still), so we won't catch every task.

Once we start collecting the data, then we can brainstorm the UI piece.

[elasticmachine]

Original comment by @bleskes:

It's important to remember that we poll every ~10 seconds (configurable, but still), so we won't catch every task.

you can work on a push model where a task that is running >1s (for example) will be monitored. Similar to slow query logs, gc issues etc.

[elasticmachine]

Original comment by @uboness:

I'd not do that for beta1. We should capture it when we figure out what we want to show. I want to avoid too many changes to the schema between releases. How about slotting this for 5.1?

[elasticmachine]

Original comment by @pickypg:

I'm never going to complain about being less aggressive with a release cycle.

[cjcenizal]

Replaced by #42621