uses some internal marked stuff, forgot about the details

now we can keep many terminals operational within the same
EtherCAT roundtrip

until now we only had one asynchronous packet on the wire.
This allows to send them as fast as possible, waiting for them
in bulk.

there is a bug in some versions of Python that asyncio.sleep(0)
cannot be cancelled.

until now this only prints whether the working counter
is wrong. For this to work we had to switch off all
fmmus that somebody else might have already set.

terminals without mailbox only have sync managers for PDOs,
those are then put where normally the mailboxes would be.

this doesn't work yet, the relays do not switch, but maybe the
terminal is broken.

this also means removing some old code and adopting terminals.
Plus, when parsing the PDOs from EEPROM, we guess from the EEPROM
entry number whether this is read or write, not from the entry
proper. At least for EL2624 this is necessary.

this unfortunately only tests without fmmu, tests with fmmu
need to be added.

we mixed up in and out pdos - at least this is what I think

add some asserts for better debugging

this was necessary for the tests, yet they still don't work

not sure whether this is the way to go, but at least it works

especially if we want to assign terminals dynamically, we cannot assing
terminals to loops at creation time.

we just try a random address, that works most of the time

there should be a test for this.

this breaks a bit the logic of terminals, but is necessary for
bootstrapping: you want to know which type of terminal it is before
we initialize it.

constantly assure that the terminals are in operational mode.

They should not drop out of operational, but if we are slow, they
do.

this is tested with read hardware, the tests obviously don't work,
we have to grandfather them.