The default hearbeat-setting is 60 seconds, so the delta between two
timeago-checks will always be >15 seconds and the timeago-alarms will
always be suspended (that's what Papertrail also shows).

To fix this, make the delta-check heartbeat-setting-dependant and also simplify
the code by using just 1 variable.

The real-world test would be to actually wait for a couple of minutes (with 2 *
heartbeat of default settings) in the unit test, but this is not feasible, so
just modify the heartbeat-setting to a lower value.

I tested it by only running tests inside `tests/timeago.test.js` and by actually
deploying this code and testing the alarms with Pushover and reading the logs in
Papertrail.

Before this change, I saw a 'Hibernation detected' log every minute. After this
change, I didn't see it anymore, probably because the app wasn't actually
hibernated (yet).

between client and server (by introducing it in sandbox.js).

On the client the behaviour is different from the server:

> On client, the issue is browsers stop the execution of JS if the window is not
> visible and the alarm is falsely triggered immediately when the execution is
> resumed, so we need to suspend the alarm for ~10 seconds after the execution has
> resumed to give the client time to update the data to prevent false alarms.

While on the server, the default heartbeat from 60s needs to be taken
into account to prevent the timeago alarm from falsely triggering.
So detect hibernation there if the last check was more than 2 heartbeats ago.

which is now required by timeago.

Also, change the timeago test a bit so that it both succeeds
when testing in isolation as when testing it along with the other tests.