🏭 Refactor metrics/events send mechanism

[kamaradclimber]

Before this patch, events and PR were sent asynchronously from callbacks.
It led to metrics not being batched (there were mostly sent one by one). It also led to loss of metrics (and sometimes fully blockage of metrics/events sending mechanism).

Now we have a queue to store pending metrics/events and both are dequeued every 10s.

This should greatly improve stability.

From the code writing perspective it requires to:

• write a callback that will be called every 10s
• from this callback schedule the execution (eagerly) of an async method
• in the async method we can finally loop over all pending metrics/events