packet-headers behavior in response to a syn flood

[pboothe]

We should make sure that packet-headers behaves no worse than the operating system with which it is surrounded in response to a SYN flood.

[robertodauria]

Some work to improve memory usage and number of goroutines has been done here: https://github.com/m-lab/dev-tracker/issues/800 (#57)

We could further improve it by adding a filter to ignore flows from IP addresses that have been sending too many SYNs in a short time, to entirely prevent spawning goroutines and allocating memory to handle them. The filtering could be done at the application level or (if this is possible at all) updating the BPF filter during capture.

[robertodauria]

Additional step in the direction of improving packet-headers' behavior under SYN floods: #58