Fix kinematic bodies not synchronizing state when using Jolt Physics

[mihe]

Fixes #101753.

Prior to #100983 we relied on a JoltBody3D::sync_state flag to indicate whether a rigid/kinematic body should have its state synchronized on the next PhysicsServer3D::flush_queries, but as of #100983 the Jolt Physics module now does the same kind of "self-listing" as Godot Physics, where each body registers with its respective physics space to have its call_queries invoked.

However, it seems I had forgot to replace one important sync_state = true with the new _enqueue_call_queries, which was the one happening in JoltBody3D::_move_kinematic, meaning we now never invoke state synchronization for things like AnimatableBody3D, which relies on this for its sync_to_physics property.

With this PR we now instead always _enqueue_call_queries if any active JoltBody3D has a valid state synchronization callback or a valid custom integration callback before stepping the simulation, which should bring it in line with Godot Physics.

(I also removed/inlined _pre_step_rigid and _pre_step_kinematic as they felt a bit excessive now.)

[rburing]

The reports_contacts() condition disappeared, is it ok? Just sanity checking.

[mihe]

The reports_contacts() condition disappeared, is it ok? Just sanity checking.

Yes! That was a remnant from the previous sync_state approach, to emulate the behavior of Godot Physics, where kinematic bodies that report contacts (i.e. a RigidBody3D frozen as kinematic) are never allowed to go to sleep and thus always have their state synchronization (and thus _integrate_forces) called.

It's not needed as of #100983, since we also prevent those types of bodies from going to sleep, meaning JoltBody3D::pre_step (and thus _enqueue_call_queries) will always be called on them.

[Repiteo]

Thanks!