OpenThread::scan is not `core::mem::forget` safe

[ivmarkov]

The gory details are described in this comment.

While not urgent, we might want to address that, because I think this is the only place in openthread which is currently not core::mem::forget safe.

[ivmarkov]

Ditto for the DNS callback.

[MabezDev]

Is this so bad? We declare in esp-hal don't use mem forget and embassy-net has similar assumptions with no actual warning at all. Hopefully Leak will land at some point and we can actually enforce it instead of suggesting, but 9/10 times the UX improvement from assuming drop can run is worth it from my experience.

[ivmarkov]

Is this so bad? We declare in esp-hal don't use mem forget and embassy-net has similar assumptions with no actual warning at all. Hopefully Leak will land at some point and we can actually enforce it instead of suggesting, but 9/10 times the UX improvement from assuming drop can run is worth it from my experience.

Well it is not the end of the world, and that's why the new DNS(-SD) API is following the simpler pre-existing scan API. For now.

With that said, scan and now dns_query are the only two such APIs, and they can be fixed by using extra memory instead of callbacks.

I'm not sure therefore - given the small surface of the core::mem::forget-unsafe APIs that we should allow this long term, specifically in openthread.

[ivmarkov]

Ditto for the new energy_scan callback.