Since each of the shelly-ht-action pods have their name and IP address
attached to the metrics collected from them, it will result in multiple
metrics existing as `last_over_time`. As a result multiple of the gauges
will show up, and things start to look silly.

This patch fixes the problem by putting an average over all values that
exist in parallel, by that dropping the instance and pod labels and
therefore result in only one metric that is displayed as intended.