Shakenbreak

----------
- Make oxidation state guessing more efficient.
- Update Quantum Espresso and FHI-aims IO functions to work with new (and old) ASE release.
- Minor updates to ensure compatibility with recent ``pymatgen`` release.
- Allow unrecognised defect names when plotting.

----------
- Add ``verbose`` option to more parsing/plotting functions for better control of output detail.
- Improve effiency & robustness of oxidation state handling.
- Miscellaneous efficiency (e.g. memory reduction) and robustness updates.
- Improved GitHub Actions test efficiency.

----------
- Add ``verbose`` options to ``io.parse_energies()`` and ``snb-parse``, also used in ``snb-plot`` and
``snb-analyse``, and set to ``False`` by default to reduce verbosity of certain SnB CLI commands.
- Use ``doped`` functions to make oxi-state guessing (and thus defect initialisation) more efficient.
- Miscellaneous efficiency and robustness updates.
- Testing updates.

----------
- ``distortion_metadata.json`` for each defect now saved to the individual defect folders (as well as the
combined total distortion metadata in the top level folder) – more likely to be retained by the user
when ``scp``\ing around etc.
- Minor updates:
- Refactor ``_format_defect_name`` to ``format_defect_name`` from ``doped`` (now a public function)
- Update ``snb-run`` to avoid possible 'file exists' warning
- Update tutorials/notebooks to specify ``vasp_nkred_std`` to streamline workflow
- Remove unnecessary ``tutorials`` folder with duplicate tutorial notebook (to reduce workload).
- Add Binder/Colab buttons to run tutorials in the cloud from docs
- Default verbosity updates (quieten some unnecessary info messages)
- Make ``distortion_metadata`` overwriting/combining more robust and less (unnecessarily) verbose
- Bugfix of ``snb-run`` from ``v3.3.0``: If max number of electronic steps (``NELM``) threshold was reached
in an ionic step, it would be falsely recognised as converged (due to ``unconverged`` being in the
``OUTCAR``). This would only affect ``snb-run`` behaviour in some cases with ``v3.3.0``, and if so the
user should be warned anyway with ``Bond_Distortion_X not fully relaxed`` when later running
``snb-parse``/``snb-plot``/``snb-groundstate``. Now fixed. To double check, one can update
``ShakeNBreak`` and just re-run ``snb-run``, and any affected distortions will be correctly determined as
unconverged and be re-submitted.

----------
- Add Dimer distortion as a targeted distortion for dimer reconstructions. It pushes two of the defect NN
to a distance of 2 Å.
- Add option ``distorted_atoms`` to the ``Distortion`` class to allow users to specify the indexes of the
atoms to distort.
- Update tests to check the new functionality.
- Update ``get_homoionic_bonds`` to detect homoionic bonds between different cations/anions (rather than
just bonds between the same element)
- Fix issue with ``snb-generate`` when no defect name was specified (by adding ``unrelaxed=True`` when
calling ``get_defect_name_from_entry``)
- Update functions that read ``OUTCARs`` to be able to read ``OUTCAR.gz`` files too
- Update energies parsing to still work when all distortions are high energy, but warn
the user about this (i.e. only ``Unperturbed``)
- Update ``snb-run`` to add early-on detection of distortions that are stuck in high energy basins and
rename them to "High_Energy" to avoid continuing their relaxation
- Miscellaneous efficiency improvements and bug fixes

----------
- Ensure the sorted ``pymatgen`` ``Structure`` is created for the VASP input (fixes a rare bug in ``v3.2.1``
and ``v3.2.2`` where for certain structures the order of elements in the POSCAR was not properly sorted,
which is usually fine, but messed with the ``ROPT`` ``INCAR`` setting).
- Plotting format updates (make legend frame more transparent to make datapoints behind it easier to see).
- Update tests
- Update docs (note about handling AFM systems)