Process perf events individually

Nobody has claimed this yet.

Hi,
after having some problems by processing individual events from a perf ring buffer (see #1894), I have discovered what it could me the root of my problem.
I have my callback function that stops and restarts the bpf program based on the events the bpf program send itself. Given this circular relationship I need to be able to control individual polls from the event buffer.

However I discovered that this is currently not possible as the poll is currently calling my callback function for all events in the buffer:
https://github.com/iovisor/bcc/blob/18d3814ca9f6bf36f3768a2d5fee0fbd55c11798/src/cc/perf_reader.c#L171-L175

Whan I need is that a call to perf_reader_poll matches to at most a single call to my callback function.
Is there a reason the current implementation processes all buffered events at once?

Start in src/cc/perf_reader.c at the poll loop linked in the issue, and trace how buffered perf events are dispatched to the callback. Confirm the requested one-event-per-poll behavior for perf_reader_poll; done means each call invokes the callback at most once while processing individual buffered events.

Written by the indexing model from the issue text.