Cached supercell orbit mappings can make structurewrangler/clustersubspace json files very large

[lbluque]

The cached dictionary for orbit mappings for different supercells can be very very large in terms of file size when saving a ClusterSubspace or a StructureWrangler with a cluster subspace that has seen many different structures and cached their orbit mappings.

This is not a "bug" per say but it can become hard to load large files since python consumes much more memory than the file itself during loading which can lead to crashing python itself.

It may be useful to find a more efficient way to save it, or an option to load or not the cached mappings.