A couple of coverity fixes

[julianoes]

Coverity sends me these emails about potential problems. In order to get to inbox-zero I had to attempt to fix some of them.

[sfuhrer]

select_airspeed_and_publish() should also be executed if no airspeed sensor is connected, so with this change I think it wouldn't work.
Inside select_airspeed_and_publish, _airspeed_validator is used yes, but only if the number of airspeed sensors is >0, in which case _airspeed_validator is != nullptr. Probably not the nicest way to program it, but should be safe, no?

[julianoes]

Aha. However, there is still one catch: if new fails to allocate because you're out of RAM _airspeed_validator is also null even though _number_of_airspeed_sensors is > 0. We need some kind of mechanism to check for that failure case and notify that airspeed is broken and we must not fly.

[dagar]

Perhaps @sfuhrer could account for the airspeed selector coverity fixes in #12887 and get the others in here?

[sfuhrer]

Perhaps @sfuhrer could account for the airspeed selector coverity fixes in #12887 and get the others in here?

Yes, I'll do that. @julianoes you can remove the 4 commits concerning the airspeed selector, I'll handle it in #12887.

[julianoes]

@sfuhrer you can change this branch and do interactive rebase as you want.

[sfuhrer]

@sfuhrer you can change this branch and do interactive rebase as you want.

Done. This PR is then ready to be merged I'd say.

[julianoes]

Thanks @sfuhrer!