First cut at EON Steel patches

[Linus Torvalds]

So I have a few more patches pending for the EON steel, and am parsing
a fair amount more data. In particular, I'm parsing things like device
ID's now, but the sad part is that I have no really good way to pass
the data back.

Why?

Part of it is an EON Steel detail: the firmware version number etc is
actually encoded per-dive: yes, you can find a "current firmware
version" for the device itself, but I'd *much* rather use the per-dive
one, which is stable (ie if I upgrade the firmware, the device
firmware version updates, but the dive info still has the firmware
version that was active for the dive). That means that the dive data
doesn't change if the firmware is updated.

But the

   device_event_emit(device, DC_EVENT_DEVINFO, &devinfo);

seems to be meant to be called just once, not once per dive. Whatever
- I guess I could just violate that unwritten rule and just do
multiple devinfo callbacks, once before each dive.

HOWEVER.

The other issue is that all of this is encoded in the same format that
all the sample data is encoded in, so it's really only available when
actively *parsing* the file. So I'd *really* like to call that
device_event_emit() when I get the set_data() in the parser - but by
then I no longer have the "device" struct to send the info to.

Can I parse the damn thing multiple times? Sure. Right now I already
parse it twice: once at parser setup time (to get the "fields" data -
dive duration, surface pressure etc), and then once for the
samples_foreach() case. Can I make the parser and the dive data
loading IO side be incestuous and parse it a *third* time early? Yes I
can, but yes I'd really prefer not to. It negates the whole principle
of separating out the dive parsing from the dive loading that
libdivecomputer seems to have.

Jef, comments?

                      Linus