Async: !Send

[bugadani]

This PR adds the inherent functions meant by #2905 to the stabilized drivers and I2S.

This PR marks Async: !Send so that async drivers can't be sent to a different core with an interrupt handler in place. Sync-ness of driver should be a separate concern that should be evaluated on a per-driver basis.

Implementations that are both Async and safe to Send can unsafely implement Send.

Closes #1525

[bugadani]

Okay I've given this PR more thought, and I think the Sendness and Syncness of driver are separate concern. We're mostly taking &mut self anyway so the Sync-ness of drivers shouldn't matter much, but they'll need to be reviewed anyway - e.g. configuring GPIOs needs some more care. Probably a lot more care.

Blocking drivers are probably fine to be Send, but they should be !Sync, so that the interrupt handlers can't cause a data race with them. We could then introduce the split interrupt handler struct along with a SyncBlocking mode, that separates the interrupt setup/handling parts from the blocking driver, and allows interrupts to be handled without locking the driver itself. I'm not convinced it's a good idea to do, but it would be possible.

[bjoernQ]

LGTM