Should IoSliceMut::advance() be used when impl std::io::Read::read_vectored()?

As of io_slice_advance is stable since 1.81, the following asserting may won't work if the users use IoSliceMut::advance in their impl of Read trait.

fn what_to_read<R>(reader: R)
where
    R: Read,
{
    let mut b1 = [0u8; 16];
    let mut b2 = [0u8; 32];

    let mut iov = [IoSliceMut::new(&mut b1), IoSliceMut::new(&mut b2)];
    let nr = reader.read_vectored(&mut iov).unwrap();

    if nr > 0 {
        assert_eq!(b1[0], iov[0][0]);
    }
    if nr >= b1.len() {
        assert_eq!(iov[0].len(), b1.len());
    }
}

It is unclear that whether the returned size of read_vectored is the real read data size, or the valid buffer size in the passed &mut [IoSliceMut<'_>].
Please clarify that so Read implementations can do the correct thing.

Start by reading the documented contracts for Read::read_vectored and IoSliceMut::advance, then compare them with the example in this issue. Clarify whether the returned count is the number of bytes read and how the passed slices may be modified, so implementations and callers have an unambiguous contract.