Add FixedBumpVec::push_in

[bluurryy]

When using a mutable BumpScope you can't have collections referring to the BumpScope at the same time. So you can't have:

struct Foo<'b, 'a> {
    bump: &'b mut BumpScope<'a>,
    vec: BumpVec<'b, 'a, u32>,
}

where vec is allocated in bump.

But you can work around it by using FixedBumpVecs and turning them into a BumpVecs temporarily:

struct Foo<'b, 'a> {
    bump: &'b mut BumpScope<'a>,
    fixed_vec: FixedBumpVec<'a, u32>,
}

impl<'b, 'a> Foo<'b, 'a> {
    fn push(&mut self, value: u32) {
        let Self { bump, fixed_vec } = self;

        let mut vec = core::mem::take(fixed_vec).into_vec(bump);
        vec.push(value);

        *fixed_vec = vec.into_fixed_vec();
    }
}

Let's add push_in(bump, value), so we can just write:

impl<'b, 'a> Foo<'b, 'a> {
    fn push(&mut self, value: u32) {
        self.fixed_vec.push_in(self.bump, value);
    }
}

[bluurryy]

This is probably a bad idea. Our api surface is already large, adding another permutation variable (*, *_try, *_in, *_try_in) should have very good reason to exist. Just some sugar like that is not it.