Overhaul stats: Refactor metrics (part 1)

[josecelano]

There is a bug in the Torrust Demo with the new metrics that I can't reproduce locally.

I will introduce some changes to the metrics package to help me find the problem.

I will also make other minor changes in that package and other packages.

Subtasks

metrics package:

• Return errors instead of panicking in the MetricCollection struct.
• Add tracing crate.
• Return an optional in get_counter_value and get_gauge_value. They are returning the default value for that metric type if the metric does not exist. That the behavior we want for increase_counter and set_gauge but not for these methods. They are currently used only in tests.

http-tracker-code, udp-tracker-core, udp-tracker-server packages:

• Add logs to listeners initialization

Change this:

pub fn run_event_listener(&mut self, receiver: Receiver<Event>) {
    let stats_repository = self.repository.clone();

    tokio::spawn(async move { dispatch_events(receiver, stats_repository).await });
}

To something like this:

pub fn run_event_listener(&mut self, receiver: Receiver<Event>) {
    let stats_repository = self.repository.clone();

    // TARGET should be HTTP_TRACKER_CORE
    tracing::info!(TARGET, "Starting event listener");

    tokio::spawn(async move {
        dispatch_events(receiver, stats_repository).await;

        tracing::info!(TARGET, "Event listener finished");
    });
}

NOTE: This won't help identify a panic inside the dispatch_events function. We should log the error returned by tokio::spawn, but the function is not async. We need to move the function call from the app setup to the app start as described here.