Real time versus normal timing

[Levansic]

I'm going to see how many writes to SD we have, and try to combine them into flattened strings from clusters or arrays. It has been nice to use the SD for prototyping and debugging, but it has been a bit too easy to arbitrarily add things to the table.

We used to think long and hard about what we would send via UDP, and there was only one, one-way stream that we would make. Now we have SD reads and writes in autonomous, teleop and timed tasks. About a quarter are no longer used, so they will get pruned. Not knowing the mechanism under SD, I just assumed that the data was coalesced and sent as part of the driver station's scheduled communication. Now, knowing that each write creates a TCP packet (pretty inefficient for a Boolean write), I have more reason to minimize those SD writes.