core: Relax lifetime constraint on msgs in I2CTransfer::transfer()

[peterdelevoryas]

This change enables users of i2cdev to create generic functions on T: I2CTransfer that have output buffers constructed separately from the I2CMessage array, like the following:

fn smbus_read_post_box<T>(i2c: &mut T, offset: u16, out: &mut [u8])
	where for<'a> T: I2CTransfer<'a>,
{
	let addr = offset.to_be_bytes();
	let mut messages = [
		T::Message::write(addr),
		T::Message::read(out),
	];
	i2c.transfer(&mut messages).expect("uh oh");
}

Before this, messages would not satisfy the constraints of .transfer(), because messages does not live as long as one of the output buffers out:

error[E0597]: `messages` does not live long enough
  --> src/smbpbisensor.rs:69:19
   |
63 |     let mut messages = [
   |         ------------ binding `messages` declared here
...
69 |         .transfer(&mut messages)
   |                   ^^^^^^^^^^^^^ borrowed value does not live long enough
...
78 | }
   | -
   | |
   | `messages` dropped here while still borrowed
   | borrow might be used here, when `messages` is dropped and runs the destructor for type `[<T as I2CTransfer<'_>>::Message; 2]`

The error message is a little confusing, but basically &'a mut [Self::Message] is forcing the array of I2CMessages to match the lifetime of the buffers in the messages, which is not strictly necessary: the array of messages can have a different lifetime than the buffers. After this change, the above example compiles successfully.

[eldruin]

Good point. Sounds good to me, thank you!
Could you add an entry to the changelog?

[peterdelevoryas]

Good point. Sounds good to me, thank you! Could you add an entry to the changelog?

Yes absolutely! Sorry, got caught up with other stuff today. Revised it just now.

[eldruin]

Thank you.
Could you move your changelog entry to the unreleased section?

[peterdelevoryas]

Thank you. Could you move your changelog entry to the unreleased section?

Oh argh, I misread the changelog section headers yeah I’ll do that

[peterdelevoryas]

Ok, changelog is now in the appropriate section

[peterdelevoryas]

CI only failed because no macOS runners are available? mebe this is ok to merge? Sorry for requiring 2 updates for the changelog :(