Optimize struct byte padding for flow.Flow

[aaronfuj]

Per #20, @cleroux noticed that we could better optimize this heavily used struct often in the hot path and being instantiated a lot.

This struct can be optimized to reduce memory usage by about 6% if my napkin math is correct. I only mention it because this seems like a struct we instantiate a lot. There are a couple gaps in this struct if we consider the 8-byte alignment.
I think you can add AppProtocol without increasing memory usage if you insert it immediately after TcpRetransmit uint32

While the AppProtocol was moved closer, this issue is to evaluate the overall struct field ordering and improve upon it to be more efficient for 64-bit architectures.

Tasks

Beta Give feedback

1. create benchmark test to assist in profiling and timing the code