More robust name validation

[aeisenbarth]

Closes #624

• This pull request changes name validation rules:
  • allow additionally . (now allowing _-. and alphanumeric, which includes 0-9a-zA-Z but also other Unicode like ɑ and ²)
  • forbid full names ., ..
  • forbid prefix __
  • forbid names only differing in character case, like abc, Abc (only one of them allowed, no matter which case)
• Name validation is now also applied to AnnData tables (keys/columns in obs, obsm, obsp, var, varm, varp, uns).
  • For obs and var dataframes, _index is forbidden.
• Validation happens at construction time when adding elements to an element type dictionary (as before).
• Additionally, validation happens before writing to Zarr.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.89%. Comparing base (8323e15) to head (726657f).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #703      +/-   ##
==========================================
+ Coverage   91.84%   91.89%   +0.04%     
==========================================
  Files          44       46       +2     
  Lines        6791     6956     +165     
==========================================
+ Hits         6237     6392     +155     
- Misses        554      564      +10     

Files with missing lines	Coverage Δ
src/spatialdata/_core/_elements.py	91.86% <100.00%> (-0.10%)	⬇️
src/spatialdata/_core/spatialdata.py	91.47% <100.00%> (+0.57%)	⬆️
src/spatialdata/_core/validation.py	100.00% <100.00%> (ø)
src/spatialdata/models/__init__.py	100.00% <100.00%> (ø)
src/spatialdata/models/models.py	87.74% <100.00%> (-0.08%)	⬇️

... and 4 files with indirect coverage changes

[LucaMarconato]

Excellent PR @aeisenbarth, thank you!

I performed my code review and applied directly the code changes. I list them here:

• I added a check for layers as well for tables (updated design docs and tests accordingly)
• There was a bug in _validate_all_elements(): it should be element_type == 'tables' (instead of 'table'`)
  • That if condition was not covered by tests, so I added a test for that
• In test_spatialdata_operations.py, some checks for tables were missing (due to old code, that was expecting a single table), I update that
• Same for some code in test_readwrite.py
• in test_writing_invalid_name() a test for labels was commented; I uncommented it
• I extended test_writing_invalid_name() to consider:
  • writing of a table with a valid name bug invalid "subnames" (this is the test I mentioned above that was not covering the "table" vs "tables" bug)
  • incremental writing of single elements (before validation of table "subnames" was triggered only on write(), now also on write_element()).
• I now trigger the name validation also on TableModel().validate() and not just on TableModel().parse(). I added tests for that in test_models.py

[LucaMarconato]

I ask you please to give a double check to my code changes, and if you agree with them (or after your edits), let's merge 😊

[LucaMarconato]

The explanation in the Discussions on how to be able to read datasets with naming problems is great! One minor todo:

• add a link to the discussion Adapting SpatialData to naming restrictions #707 in the exception that the code raises when reading a dataset with naming problems.