Suggestion: Implement into_boxed_slice() for TinyVec

[2moe]

While we can currently convert TinyVec to Box<[T]> via tinyvec.to_vec().into_boxed_slice(), this incurs unnecessary memory reallocation.

Implementing tinyvec.into_boxed_slice() would be better.

Since Box<[T]> is more compact and emphasizes immutable length, it is useful in certain situations.

use core::mem::size_of_val;

use tinyvec::{Array, TinyVec};

struct Foo {
    bar: Box<[u32]>,
}

fn main() {
    // Initialize TinyVec with 204800 elements (exceeding inline capacity)
    let tv: TinyVec<[_; 102400]> = (0..204800).collect();
    assert!(tv.is_heap());
    assert_eq!(size_of_val(&tv), 409608);

    let bar = tinyvec_into_boxed_slice(tv);
    assert_eq!(size_of_val(&bar), 16); // Final size: 16 bytes (fat pointer)

    let foo = Foo { bar };
    assert_eq!(foo.bar.get(..3), Some(&[0, 1, 2][..]));
}

fn tinyvec_into_boxed_slice<T: Array>(
    owned: TinyVec<T>,
) -> Box<[<T as Array>::Item]> {
    use TinyVec::*;
    match owned {
        Heap(inner) => inner,
        Inline(mut inner) => inner.drain_to_vec(),
    }
    .into_boxed_slice()
}

[Lokathor]

Sounds great. Do the PR and I can probably review this pretty quick.