this is a simplification in preparation for the process-based
EtherCat devices. It seems to work, why did it use to be more
complicated?

also improve error messages on the way

if the loop is already running, we do not want to re-initialize the
terminals, this would stop it.

we returned two different things whether we were keeping logs
upon loading or not. This makes it hard to program around.

Now only return logs, and store the file descriptor internally.

this seems to be a subtle change in recent versions of Python

this will hopefully make a fan-in possible: the EBPF changes the
protocol on the incoming packet, and we bind to this protocol.
This way two processes may listen to different subsets of ethercat
packets.

this also adds support for terminals which have both
PacketVars and ProcessVars

this uses less cool features of asyncio, but does actually what
we want.

show what is bad, without overloading the output

newer versions of Python do not allow binding of sockets 'by hand'
anymore. Interestingly, everything works perfectly without. Was
it ever needed? Not sure.

now terminals can be initialized with either absolute or relative only,
the other one being chosen sensibly. Also remove race condition when
searching for free addresses.

this way we can read individual object entries without reading
the entire list.

and also improve its output a bit