As the RFFE hardware changed, we did not update
the bpm-sw communication with it. Many variables
have ceased to exist and other have been included.

The correct place t destroy it is in
eth_release (), as we assign it in
eth_open ()

This fixes the issue where we want just to read
the last channel acquisition. Before this patch,
we would always read the trigger address from
the last acquisition independent of the channel.

This would result in wrong data, if the last channel
was not the same as the current.

This is documented on github issue #119.

On BPM server, even if a field is not used, trhe client must send
it. So, that the server can correctly validate it size.

Now, it's possible to select a range of channels to apply
the same configurations.

Github commit lnls-dig/bpm-gw@d952234374 changed it.
So we update it here.

This is good, as if there is a timeout on the PCIe
side for some reason not detected (short reading),
on a next llio initialization, it will be reset.

According to AD9510 datasheet this is not needed on this
register only. So, we removed it here.

It didn't add any value to debug. So, we removed it here.

Sometimes, only 30ms is not enough to si57x to reset
(maybe we are a bit oo fast on waiting). So, we use 300ms
and everything worked again.

We ned to make sure we have incoming clock to
ISLA216P ADCs. Otherwise we can't initialize
it with certainty.

This fixes #118.

(cherry picked from commit d174d495)

The previous function was wrong, as it was
configuring trigger iface and not trigger mux.

Added "_BASE" to name address, for consistency
with existing names.