This allows these methods to be called from ractors.

Add new exported function `rb_vm_lock_with_barrier()` that requires
users to include "vm_sync.h"

The embed layout is way more common than the heap one,
especially since WVA.

I think it makes for more readable code to inverse the
flag.

`RSHAPE_PARENT` is error prone because it returns a raw untagged
shape_id.

To check if a shape is a direct parent of another, tags should be
discarded. So providing a comparison function is better than exposing
untagged ids.

Followup: https://github.com/ruby/ruby/pull/13589

This simplify a lot of things, as we no longer need to manually
manage the memory, we can use the Read-Copy-Update pattern and
avoid numerous race conditions.

Co-Authored-By: Étienne Barrié <etienne.barrie@gmail.com>

The type isn't opaque because Ruby isn't often compiled with LTO,
so for optimization purpose it's better to allow as much inlining
as possible.

However ideally only `shape.c` and `shape.h` should deal with
the actual struct, and everything else should just deal with opaque
`shape_id_t`.

Now that we have the `heap_index` in shape flags we no longer
need `T_OBJECT` shapes.

Instead it's now a `shape_id` flag.

This allows to check if an object is complex without having
to chase the `rb_shape_t` pointer.

Instead `shape_id_t` higher bits contain flags, and the first one
tells whether the shape is frozen.

This has multiple benefits:
  - Can check if a shape is frozen with a single bit check instead of
    dereferencing a pointer.
  - Guarantees it is always possible to transition to frozen.
  - This allow reclaiming `FL_FREEZE` (not done yet).

The downside is you have to be careful to preserve these flags
when transitioning.

We should avoid conversions from `rb_shape_t *` into `shape_id_t`
outside of `shape.c` as the short term goal is to have `shape_id_t`
contain tags.