Implement data collection to send to monitoring

[alexgarel]

After [identification of metrics](see #661) we can implement collecting those metrics.

To avoid putting postgresql database under too much stress we would prefer to collect info in the monitoring influxdb database, using a recurring task in scheduler.

Hourly granularity might be enough.

Ideally we should remember last time we computed the metric, to be able to fill the gap if the metric was stalled for some time (maybe add a table to store when each task was last run). This could also be a way to bootstrap a new metric and add previous data that can be computed.