Add metrics for trace event perf event monitor

[christos68k]

I noticed when working on #307 that we don't have metrics for reading trace events from the perf event buffers (we collect them but don't use/report them) which means that we don't know how often we drop events.

The current trace event processing scheme is synchronous with reading events from the perf events buffers (we don't read events in a batch and then process that batch, we read a single event, process that event, read next event, ..) which means that we could be dropping events if processing can't keep up.