7 message read write

[cbaxter]

Completed initial Serialization support. Was able to keep latency down to 10/15 [us] range for NamedPipes and TCP respectively.

Further optimization possible with sockets (separate issues created).

Not entirely happy with FrameReader code; but makes a notable boost to performance.

A couple of thoughts:

• Removing FramedOutputStream greatly improved performance; only drawback is entire message serialized in memory before being transmitted over the wire. Really should only be an issue if sending very large messages (which wouldn't be recommended regardless).
• Several issues created to handle further improvements.

[jgoz]

Looks good to me.

As discussed, we can try to improve the Type header performance and size, possibly on a per-serializer basis. I'll create a separate issue for this.