Replace existing redundant congestion control mech with reactor specific limitations

[evan-forbes]

Currently, the p2p relies heavily on a per peer rate limiting the send and receive rate. Peers can send redundant messages or requests, and they are not kicked. Instead, they are free to absorb the entire allocated bandwidth without reprocussion.

This is very problematic, as nodes must keep the allocated bandwidth low, which results in two nodes artificially limiting bandwidth they otherwise could be utilizing.

The more time consuming approach is to visit each and every message that is sent or received, and ensuring that there are appropriate protections in place for that message. In general, this will likely fall into these catagories:

• stateful protections (protect against redundant or unexpected messages)
• stateless validation (already completed)
• rate limits on heavy, non-critical, pull based logic (essentially just blocksync)

To close this issue, we need to add these checks for each message in each reactor. Some new infrastructure might be required for rate limits and redudant message checks.