Packet queue performance can be improved

[nhuang-tt]

The base class of the routing kernels, packet_queue, has a lot of branches and accesses data members at runtime that can be otherwise done at compile time. Just running test_tx_rx (Traffic generator to receiver bandwidth test), the baseline is 0.84 B/cycle and wall clock bandwidth is 22.86 MB/s. After changing the if statements to compile time it increased to 0.95 B/cycle and 26.45 MB/s.

[davorchap]

@ubcheema sync with Nigel

[nhuang-tt]

Yes, I will make the updates after the dependency: power of 2 ring buffer size ticket