tcp nodelay to avoid batching TCP packets, handle multi-frame messages/packets

[davidt315]

Fix 1: Add the TCP Nodelay- batching the TCP packets with the default behavior of the TCP socket resulted in many dropped packets (see comment).

Fix 2: Handle multiple protocol messages in the TCP buffer - the driver's receive callback on the TCP stream made an assumption that there would only be a single AXIO protocol message within a single receive buffer from the TCP message. After this fix, it iterates through the TCP buffer to find any additional AXIO prefixed protocol messages

Fix 3: Handle multiple CAN Frames within a single AXIO Protocol message. Each AXIO prefixed protocol message can have more than a single CAN Frame packed inside of it. The previous iteration of the driver ignored any frames beyond the first one. This fix iterates through all the possible CAN Frames

Potential Fix 4: Increase the buffer size to prevent any chance of overflow on the receive buffer reaching the end. Not observed but a cautious forward looking fix. Also sets the buffer once as a member variable instead of creating a new buffer on the heap per-receive for minor efficiency boost

Known Remaining Issues:

1. Not handling axiomatic-generated heartbeats or sending heartbeats to the axiomatic to keep connection live, which has not been a problem as they are silently skipped and TCP doesn't require sending the heartbeats from the documentation
2. Not handling axiomatic status messages or CAN FD streams, which has not been a problem in current usage
3. No UDP support

[davidt315]

With fix:

automat@robot01:~$ python3 ~/can_compare.py --duration 10 --gap 1 --rx-iface can0 --tx-iface vcan0 --data DEDEDEDEDEDEDEDE --load 90
Output directory: /tmp/can_compare
Duration:         10s
Senders:          4x  (estimated load 80%)

Running... 1s remaining  
                        Draining pipeline... done
                                                 WARNING: bus load 100.2% exceeded 95% — results may be affected by TX queue overflow.

Results
  Sent    (vcan0):  37607
  Received (can0): 37607
  Delta:                    0
  Loss:                     0.00%
  Bus load:                 100.2%

Inter-frame timing
  TX candump (vcan0): avg=0.266ms  min=-0.002ms  max=1.067ms
  RX candump (can0): avg=0.266ms  min=0.098ms  max=0.751ms
  pcap       (br0): avg=0.308ms  min=0.008ms  max=0.998ms

Data comparison
  Corrupted frames (same position, different content): 0
  All frame data matches.

Output files:
  TX candump: /tmp/can_compare/candump_vcan0.log
  RX candump: /tmp/can_compare/candump_can0.log
  pcap:       /tmp/can_compare/axiomatic_capture.pcap

Before Fix:

automat@robot01:~$ python3 ~/can_compare.py --duration 10 --gap 1 --rx-iface can0 --tx-iface vcan0 --data DEDEDEDEDEDEDEDE --load 60
Output directory: /tmp/can_compare
Duration:         10s
Senders:          3x  (estimated load 60%)

Running... 1s remaining  
                        Draining pipeline... done

Results
  Sent    (vcan0):  28178
  Received (can0): 26965
  Delta:                    1213
  Loss:                     4.30%
  Bus load:                 75.1%

Inter-frame timing
  TX candump (vcan0): avg=0.355ms  min=-0.006ms  max=1.073ms
  RX candump (can0): avg=0.377ms  min=0.125ms  max=148.078ms
  pcap       (br0): avg=12.805ms  min=0.014ms  max=243.582ms
  note: pcap shows TCP segment intervals — bridge is batching ~4.8 frames/segment (779 segments, 3771 frames)

Data comparison
  Corrupted frames (same position, different content): 0
  All frame data matches.

Output files:
  TX candump: /tmp/can_compare/candump_vcan0.log
  RX candump: /tmp/can_compare/candump_can0.log
  pcap:       /tmp/can_compare/axiomatic_capture.pcap