CNET Stack per Channel instead of per Graph

[jalalmostafa]

The current implementation binds a CNET stack to a specific thread and CNET graph. This limits the operations of a channel to a particular thread and then we cannot move channels among threads.
I have a suggestion to bind the stack to the channel itself, the advantages of this approach:

• Flexibility to execute channels on any thread as it state is owned by itself not by the thread
• Less latency: There is no need to look up the stack to find the channel PCB because it has a direct pointer to its PCB entry.
• Can minimize shared data in CNET graphs, and open doors to multithreading in CNET graphs
• Can open doors to use thread pool rather than dedicated threads.

Disadvantage:

• A major change in CNDP, I do not have other disadvantages in mind.

Comment: Channels should stay accessed by only 1 thread at a time regardless of how many threads there are.