Protocol framework: use pooled ByteBuf allocator

This should enhance the default performance characteristics by reusing
ByteBufs if possible. With 4.0.19, netty keeps a thread-local cache of
direct buffers, so we should see a reduction in buffer copying, if not
other effects.

Change-Id: I81fcf380f6f3b03187000dbca0dd8507975e8dd1
Signed-off-by: Robert Varga <rovarga@cisco.com>