Working publisher using broadcast

[Carter12s]

Description

Re-working publisher to use tokio::broadcast as the underlying channel so that one lagging TCP Socket doesn't result in the publisher blocking.

Fixes

Closes: #206

Checklist

• Update CHANGELOG.md

[lucasw]

I can test this more in a few days- but did I try removing the publish async: lucasw@395cf99 - want to have that change here (and then deal with all the downstream pubilsh() changes)?

[Carter12s]

I'd like to avoid the removing async from publish change in this MR. I don't think it is a crazy change to make, but I think one change at a time.

[lucasw]

The intended fix here looks good, and the connection even recovers in the case of disconnecting ethernet, waiting a minute, then reconnecting (but too much longer gets more fully disconnected or it's just hung up forever and even reconnections don't recover it?).

[Carter12s]

I went ahead and merged with your testing cause I like these changes, but definitely more work to do here to make ros1 robust over various disconnection scenarios. Lots of complexity buried in there.