Poi label names

[Hendrik-code]

Summary

Overhauls POI naming across POI and POI_Global per the requested spec. Three focused commits; full unit suite 406 passed / 4 skipped (395 baseline + 11 new POI tests). Coloring/serialization of .mrk.json is kept byte-identical.

What changed (mapped to the requirements)

1. Names survive all POI ops — POI.copy, _set_inplace, POI_Global.to_other, calc_poi_average, and POI_Descriptor.copy now carry level_one_info/level_two_info/info/definition.
   Since reorient/rescale/extract/map_labels/to_global/average route through copy(), the names follow.
2. Index by level names — poi[idx, "level2name"] and poi["level1name", "level2name"] work for get/set/in (via unpack_poi_id → level_*_info._get_id → descriptor definition), alongside ids and Enums.
3. Enums everywhere + warning — Enum keys use .value; a warning fires when the matching level_*_info is still Any (unset).
4. poi.level_one_name(region) — returns the group (level-one) name.
5. poi.label_name(region, subregion) — returns the point name. set_label_name/set_level_one_name write them.
6. New nested label_name format — {region: {subregion: name, "name": group}} instead of flat {"(1, 2)": "C2"}. normalize_label_name() migrates the old flat form (and JSON string keys) on load;
   save always writes the new form. .mrk.json export/import stays byte-identical, now fed from the nested dict. join_left/txt-import updated.
7. Priority + warning — a custom label_name wins over the level_x_info enum name, with a warning when it diverges from a given enum name.

Notes for reviewers

• Location.save_as_name() is False (subregions saved as ints), so its _get_name returns the id string — the accessor treats that as "no enum name" and falls back to the id/custom name. Vertebra_Instance uses real names.
  (Tell me if you'd like Location to also resolve to enum names.)
• Name indexing is wired into __getitem__/__setitem__/__contains__; extract_region/extract_subregion/remove still take ids+Enums (not yet level names) — easy to extend.
• Backwards compatible: old flat-format POI files load and migrate transparently; adds unit_tests/test_poi_label_names.py (11 tests) locking migration, accessors, round-trip, old-format load, name indexing, warnings, and metadata transfer.

Commits:

• 18d9096 fix(poi): carry level_one_info/level_two_info/info through all POI ops
• 38be00e feat(poi): nested label_name format + name accessors + migration
• 3ab343d feat(poi): index POIs by level names + warn on enum without level_*_info

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[robert-graf]

Do not merge for the new release. Needs a lot of testing.