This, together with commits from c171576c
to bbe05bf6 fixes #115 github issue

In this way, we reduce code repetition and
maintenance becomes easier

In this way, we reduce code repetition and
maintenance becomes easier

This opens new possibilities for chaining SMIOs.
For instance, the FMC_250M SMIO can register 2
more SMIOs: FMC_ADC_COMMON and FMC_ACTIVE_CLK

This was include in commits from lnls-bpm/bpm-gw@09f35bd8f7 to
lnls-bpm/bpm-gw@3c537b769f

The FPGA gateware was changed to uniformize
the WB register map for the FMC ACTIVE part.

Now, if EPICS IOCs are started with systemd
they have a dependency with bpm-sw@.service

Removing the socket handler on *_cfg handle was wrong
as the devio_destroy_smio () already calls
devio_destroy_actor (), which, in turn, removes
the socket from the reader.

Also, as we expect only a single message on the
handler, it`s safer and cleaner to not poll the socket
until there is no message.

If an SMIO fails for some reason, continue to spawn
the other ones.

This was changed on commit 5a769af3

As an optimization, we registered the SMIO message
and SMIO config handlers before creating the actor.

For the first case this is OK, as as register the
handler to an existing socket creted by
zsys_create_pipe ().

For the second case the registered socket does not
yet exist, as it will be creted by zactor_new ().

This fixes #114 github issue

It's easier to standardize the initial EPICS
BPM ID from 1 to 24 instead of from 0 to 23.

In this way, the reported ID to clients stau consistent
with the FPGA driver, as well.

In this way, we avoid creating an SMIO instance
without a listener on the other side.