Add naming scheme selection to run configuration

Add a nullable naming_scheme column to NuzlockeRun so users can pick a
themed word category for nickname suggestions. Includes Alembic migration,
updated Pydantic schemas, a GET /runs/naming-categories endpoint backed by
a cached dictionary loader, and frontend dropdowns in both the NewRun
creation flow and the RunDashboard for mid-run changes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.beans/nuzlocke-tracker-ueyy--create-name-dictionary-data-file.md

@@ -5,25 +5,34 @@ status: todo
 type: task
 priority: normal
 created_at: 2026-02-11T15:56:26Z
-updated_at: 2026-02-11T15:56:48Z
+updated_at: 2026-02-11T20:23:29Z
 parent: nuzlocke-tracker-igl3
 blocking:
     - nuzlocke-tracker-c6ly
 ---
 
-Create a JSON data file containing themed words for nickname generation.
+Create a JSON data file containing themed words for nickname generation, stored in the backend alongside other seed data.
 
 ## Requirements
 
-- Each word entry has: `word` (string) and `categories` (string array)
+- Store at `backend/src/app/seeds/data/name_dictionary.json`
+- Category-keyed structure for fast lookup:
+  ```json
+  {
+    "mythology": ["Apollo", "Athena", "Loki", ...],
+    "space": ["Apollo", "Nova", "Nebula", ...],
+    "food": ["Basil", "Sage", "Pepper", ...]
+  }
+  ```
+- Words may appear in multiple categories
 - Categories should include themes like: mythology, food, space, nature, warriors, music, literature, gems, ocean, weather, etc.
 - Target 150-200 words per category
-- Words can belong to multiple categories
 - Words should be short, punchy, and suitable as Pokemon nicknames (ideally 1-2 words, max ~12 characters)
+- This file is NOT seeded into the database — it is read directly by the backend service at runtime
 
 ## Checklist
 
-- [ ] Define the data schema / TypeScript type for dictionary entries
-- [ ] Create the JSON data file with initial categories
+- [ ] Create `backend/src/app/seeds/data/name_dictionary.json` with the category-keyed structure
 - [ ] Populate each category with 150-200 words
-- [ ] Validate no duplicates exist within the file
+- [ ] Validate no duplicates exist within a single category
+- [ ] Add a utility function to load the dictionary from disk (with caching)