Data Refactor

[tylerflex]

Made a new branch to start the data refactor from scratch.

To Do List:

• reorganize N2F: move near2far.py to components/, merge monitor_data.py and `monitor_data_n2f.py
• clean up imports: move all td namespace imports to tidy3d/__init__.py, move Symmetry to types.py.
• simplify file IO: exclude data from .json, write components as string in hdf5.
• add frozen data fields: replace MonitorData methods (eg. apply_symmery() with properties using stored).
• Simplify DataArray subclassing from xr.DataArray
• introduce Dataset abstraction layer, test if circular imports are avoided.
• fix sim_data: make monitor_data a property and add tuple of MonitorData, fix the getitem and other operations.
• improve test coverage for data components.
• make updater work with new changes.
• fix up backend: add fields into the MonitorData constructors.
• final clean up / documentation

Backend Changes Needed

• when the data is empty and constructing the DataArray on the backend, use np.zeros() with the correct shape (including a 0).
• need to pass symmetry, symmetry_center, and grid_expanded to FieldData -like objects when they are created. If left out, it will just assume symmetry=(0,0,0).
• there is no longer a .load_from_handle() classmethod, it’s used a couple times in the backend, we need to find a different approach.
• the coords must be aligned with the shape of the DataArray.data. Sometimes this might not be the case, I had to fix a few times in the front end.