Modifications for network stack performance testing

[martinkarsten]

While neper is primarily a networking benchmark tool, it looks like it is also often used for network stack performance testing. These are various suggestions for improvements for this use case:

'Optimistic' I/O: A lot of applications used this pattern by which a server does not check epoll first before writing a response to a socket.

'Discard': For some reason, it seems the first (1,N?) measurements in the RR experiments are way off. Not sure whether those are just cold-start effects or something else is amiss. However, it does distort the latency measurements.

'Busywait': This mode allows using busywaiting across epoll and napi busypolling. A special case (0 < N < 1000) can be used for epoll busywaiting without napi busypolling.

The Makefile changes are just for good measure.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.