Move emit of primary-GPS-changed into AP_GPS

[peterbarker]

This wasn't really a sensor-health thing, either....

New addition to the blending test ensures we see the message come out in the log

This change will mean all vehicles can expect to see this event emitted.

[amilcarlucas]

Nice change

[rmackay9]

As discussed, let's initialise the primary value so we don't get a message in the log every time soon after startup.

[peterbarker]

Incidentally, I'm kind of the opinion it would be OK to retain the current behaviour of logging a primary change if you're set up to not-auto-switch and not auto-switching. It's an atypical setup and we could save flash and complexity.

[peterbarker]

I've added the requested change - so now we don't emit this message for any change on the first loop, which is what will happen if you both set GPS_PRIMARY to (e.g.) 1 and change GPS_AUTO_SWITCH to zero.

Initialising the primary_instance state to be equal to the GPS_PRIMARY variable is insufficient without also checking GPS_AUTO_SWITCH. While we probably should honour GPS_PRIMARY even at initialisation regardless of GPS_AUTO_SWITCH being on or off, this patch should still be the same shape IMO - leave the logic in update_primary, don't duplicate it in initialisation.

[peterbarker]

Oh. Tested in SITL by setting LOG_DISARMED, GPS_PRIMARY and GPS_AUTO_SWITCH and checking what EV messages got emitted.