[Tracking issue] p2p connection optimizations

[cason]

Channel multiplexing:

• p2p: connection multiplexing does not work as expected on the receiving side #3250
• p2p: Sending a high-priority msg can take a long time due to full MConnection queue #2128
• Make channel prioritization based on number of messages in channel (vs number of bytes sent) #2954

Connection's performance:

• Speedup CPU time of channel.WritePacketMsgTo #2951
• Change flowrate.clock() to be based on a variable that is scheduled to update every 20ms #2958
• Speedup the UpdateStats window #2956
• Lower default flush timeout #2988
• Secret Connection: Remove sync.Pool, and put those buffers in the connection struct #3197
• Secret Connection: Buffer the underlying connection #3198

Marshalling:

• Make reactor.onReceive not block unmarshalling more packets from peer #3199
• p2p: add a SendMarshalled or SendBytes method to p2p.Peer interface #3185
• Make reactor.onReceive not block unmarshalling more packets from peer #3199
• p2p/conn: Why do we use protobufs for low-level communication packets? #3469

Other:

• Backport "config: add bootstrap peers" (#9680) to v0.37.x #1016
• p2p: dialing addresses provided by a seed node disregards MaxNumOutboundPeers parameter  #486
• Lower allocation overhead of metrics in peer.Send #2840
• perf(consensus): Run broadcast routines out of process #3180
• p2p: investigate why re-dialing persistent peers consumes so many resources #3267
• Stop overwriting msgBytes when receiving data #3237
• p2p: forever connected to lazy peer with no progress #3316
• p2p: add HasChannel method to p2p.Peer interface #3472
• p2p: remove redundant checks on sending methods #3511

Documentation/spec:

• spec/p2p: Document channels and multiplex connections #600
• p2p: review and update the documentation of the p2p configuration parameters #1038

[cason]

Not part of, but related: #2955

[cason]

Not part, but related: #2685