Fix spglib >= 2.5 compatibility: convert SpglibDataset to plain dict

[abhinavships]

# Fix spglib >= 2.5 compatibility: SpglibDataset dataclass dict access

Closes #332 — "An update with spglib"

---

Problem

spglib >= 2.5.0 changed get_symmetry_dataset() to return a typed
SpglibDataset dataclass object instead of a plain Python dict.

JARVIS-Tools accesses the result throughout Spacegroup3D using
square-bracket (dict-style) syntax, for example:

dataset["international"]   # TypeError on spglib >= 2.5
dataset["number"]
dataset["wyckoffs"]

This raises TypeError: 'SpglibDataset' object is not subscriptable
for anyone running a modern spglib installation, breaking spacegroup
determination, CFID descriptors, and any downstream workflow that calls
Spacegroup3D.

---

Fix

Added a single small helper function _dataset_to_dict() at module
level (≈ 25 lines):

def _dataset_to_dict(dataset):
    if dataset is None:
        return None
    if isinstance(dataset, dict):      # spglib < 2.5 – already a dict
        return dataset
    import dataclasses
    if dataclasses.is_dataclass(dataset):   # spglib >= 2.5 dataclass
        return dataclasses.asdict(dataset)
    return dict(dataset)               # fallback for future changes

Then called it in exactly two places inside Spacegroup3D:

1. __init__ — normalise the dataset argument before storing it.
2. spacegroup_data() — normalise the return value of
   spglib.get_symmetry_dataset() before constructing a new
   Spacegroup3D.

No other code in the class was changed. The fix is fully backwards
compatible — a plain dict passes through _dataset_to_dict()
unchanged, so users on older spglib versions are unaffected.

---

Testing

Verified locally with both spglib 1.16.5 and spglib 2.5.2:

coverage run -m pytest jarvis/tests/testfiles/analysis/structure/

All existing tests pass on both versions.

---

Files changed

• jarvis/analysis/structure/spacegroup.py — added _dataset_to_dict();
  patched __init__ and spacegroup_data().