feat(state): make pbc a per-system attribute

[lil-lon]

Summary

• Make SimState.pbc a per-system attribute of shape (n_systems, 3); a bool, list[bool], or (3,) tensor is broadcast to all systems. Moving it from _global_attributes to _system_attributes lets slicing, splitting, popping, and concatenation handle it like cell.
• atoms_to_state / structures_to_state accept inputs with different pbc instead of raising, and state_to_atoms / state_to_structures / trajectory files store each system's own pbc row.
• Ewald and PME models evaluate the periodic systems of a mixed batch and zero-fill systems with no periodic axis.
• Add require_full_pbc; elastic tensor calculation now rejects systems that are not fully periodic.
• The orb / NequIP / metatomic wrappers convert pbc back to the legacy (3,) row on a shallow copy (_to_legacy_pbc_state) before delegating, because those integrations broadcast state.pbc themselves assuming the old shape; mixed-pbc batches raise a clear error there.
• Declare pbc after system_idx so __setattr__ can broadcast a (3,) row already during __init__, leaving a single broadcast site (__post_init__ only validates the shape). Simplify pbc handling in autobatching and io now that pbc is always a tensor.

Related issue: #584

Compatibility

Breaking:

• External code reading state.pbc as a global (3,) tensor must index it per system. The bundled orb / NequIP / metatomic wrappers already adjusted this.
• get_attrs_for_scope: pbc moved from "global" to "per-system".

Unchanged:

• bool / list[bool] / (3,) inputs are still accepted and broadcast to all systems.
• Trajectory files still store a (3,) pbc array; existing files stay readable.

Discussion points

• _to_legacy_pbc_state hands third-party forward() a SimState whose pbc temporarily violates the (n_systems, 3) invariant. The preferred fix is upstream support for per-system pbc in the integrations; until then the shim keeps the current orb / NequIP / metatomic releases working unchanged, since they only read plain attributes off the state.
• state.pbc = <1-D value> after construction broadcasts any length without validation (__post_init__ runs only at construction). Missing validation is pre-existing, but should __setattr__ enforce a length-3 row?

Checklist

Before a pull request can be merged, the following items must be checked:

• Doc strings have been added in the Google docstring format.
• Run ruff on your code.
• Tests have been added for any new functionality or bug fixes.