As discovered in https://gitlab.com/gitlab-org/gitlab/-/issues/231248
the `GlobalSearchSampler` which invokes this
`Elastic::MetricsUpdateService` is using an extra DB connection since
it runs in it's own thread. This extra DB connection is not accounted
for and can cause issues with the connection pool.

We could account for this extra connection but it turns out the 2
database queries happening in here aren't really necessary anyway. The
first checks if the Elasticsearch integration is enabled. Skipping this
is safe since it will just set `0` for these metrics which is not a
problem. The 2nd checks to see if prometheus monitoring is enabled. But
this check was already happening before spawning the
`GlobalSearchSampler` in
[`config/initializers/7_prometheus_metrics.rb`](
https://gitlab.com/gitlab-org/gitlab/-/blob/17ab3d1634f5f1ca82334f270cf8239d7170b0a2/config/initializers/7_prometheus_metrics.rb#L41
) so we can safely skip that too.

The only other things happening now in this class are requests to Redis
and don't require a DB connection.