i have a tuyamcu based device that occasionally gets a flipped bit in
messages it receives from the muc. those usually show up as checksum
failures, but if the bit flips in the length field then we wait for
bytes that just arent going to arrive, so we don't get to the cksum
field for that test to fail. instead, add a timeout that the tick
checks, and reset the recv state machine on a timeout.
if the message that was corrupted was a dp update, we'll end up with
an inconsistent view of the state of the DPs. maybe we should send a
request for all datapoint values when this or a cksum failure happens?
052b5c2926