Make it possible to view high cardinality metrics over longer timeframes

Part of #2026 (closed).

A new table has been added to a dashboard via the MR gitlab-com/runbooks!5264 (merged), with the endpoints sorted by traffic share. However, as of now, it's not feasible to display the table for time frames bigger than 2 days due to the amount of timeseries data to be queried. It would be more accurate to visualize this data over periods of 1 week or more since daily or hourly does not represent the biggest offenders accurately.

We currently record the gitlab_application_sli:* aggregations to view some of the SLIs that feed into the error budgets for stage groups, but because of the number of series (endpoints, sidekiq workers, ...) the cardinality is still quite high. Making it impossible to load more than a couple of days of data.