WANProxy I/O performance improvements.

[Mallett, Juli]

Just wanted to give a heads-up to the list that some structural
changes that I've been thinking about for some time to the I/O and
buffer systems have finally gone in to WANProxy.  I doubt that these
are the bottlenecks in any real-world deployments, but I figure every
little bit counts.  There's actually some slowdown in the XCodec as a
result, but I think it's less major than the I/O throughput gain.

(For some idea of what I'm talking about, the benchmark in
io/example/read-sink1 which reads data from stdin as fast as it can
has gone from being able to sustain about 150MB/s on one of my systems
to now being able to sustain on the order of 1GB/s.)

Feature improvements are still on my roadmap, but it seems like I'll
have more chances to improve infrastructure (e.g. multi-threading
support) before I'll have a chance to work on any major features.  I
find it easier to work on infrastructural changes over a long period
of time in small amounts than features, and my dayjob hasn't been
leaving me a lot of time for WANProxy.

Thanks,
Juli.