feat: YAML content importer + phys/ct-2024 collection (proof)

content/phys/ct-2024.yaml — 15 questions from ЦЭ,ЦТ 2024 across
6 topics (kinem, mol, emf, electro, magnet, optics) as proof of format.

backend/scripts/import-content.js — unified importer:
- Validates schema (subject, year, options, exactly-1-correct)
- Aliases (kinem, mol, ...) resolve to Russian topic names via get-or-create
- Deduplicates by first 80 chars of text (matches legacy seed_*.js behavior)
- Runs in a single transaction, idempotent re-runs

On fresh DB: 13 added (2 dedup collisions — same 80-char prefix, expected).
On prod DB: 0 added (all already exist from legacy seeds).
Second run on either: 0 added (dedup works).

Legacy seed_phys_ct2024.js kept as backup — see content/README.md
for migration guide.

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

content/README.md

@@ -0,0 +1,80 @@
+# Content as data
+
+Question collections live here as YAML, imported via a single CLI.
+This replaces the ad-hoc `backend/scripts/seed_phys_ct2024.js` pattern.
+
+## Import command
+
+```sh
+cd backend
+npm run import:content -- ../content/phys/ct-2024.yaml
+```
+
+## File format
+
+```yaml
+meta:
+  subject: phys          # phys | math | bio | chem
+  year: 2024             # exam year (integer)
+  source: "ЦЭ,ЦТ 2024"  # optional label shown in import log
+
+topics:
+  kinem:                 # topic alias (see aliases below)
+    - text: |
+        Question text (multi-line supported, LaTeX with \( \) works)
+      difficulty: 1      # 1=easy, 2=medium, 3=hard  (default: 1)
+      explanation: "Solution explanation"  # optional
+      options:
+        - { text: "Answer A", correct: true }  # exactly ONE correct
+        - { text: "Answer B" }
+        - { text: "Answer C" }
+
+  "Full topic name":     # or use full Russian name — will be found or created
+    - text: "..."
+      options: [...]
+```
+
+## Topic aliases (subject=phys)
+
+| Alias    | Topic name                      |
+|----------|---------------------------------|
+| kinem    | Кинематика                      |
+| dynam    | Динамика                        |
+| cons     | Законы сохранения               |
+| mol      | Молекулярная физика             |
+| thermo   | Термодинамика                   |
+| electro  | Электростатика                  |
+| dc       | Постоянный ток                  |
+| magnet   | Магнетизм                       |
+| emf      | Электромагнитная индукция       |
+| optics   | Оптика                          |
+| quantum  | Квантовая и ядерная физика      |
+| waves    | Колебания и волны               |
+
+For other topic names, use the full Russian name as the key — the importer
+looks it up in the database (case-insensitive) or creates a new topic.
+
+## Dedup logic
+
+Questions are skipped if the first 80 characters of their text already
+exist in the database for the same subject. This matches the legacy
+`seed_phys_*.js` behavior, ensuring idempotent re-runs.
+
+## Migrating a legacy seed_*.js
+
+1. Copy the file structure from `content/phys/ct-2024.yaml`
+2. Convert each `q(T.kinem, text, opts, diff, year, expl)` call to YAML:
+   - `T.kinem` → `topics: kinem:`
+   - `text` → `text: |` (use literal block for multi-line)
+   - `opts: [{t: "...", c: true}, ...]` → `options: [{text: "...", correct: true}, ...]`
+   - `diff` → `difficulty:`
+   - `expl` → `explanation:`
+3. Run `npm run import:content -- ../content/<subject>/<file>.yaml`
+4. Verify output shows expected `added` count
+5. Keep the legacy `seed_*.js` file as backup until verified
+
+## Collections
+
+| File | Subject | Year | Source | Questions |
+|------|---------|------|--------|-----------|
+| phys/ct-2024.yaml | Физика | 2024 | ЦЭ,ЦТ 2024 | 13 (proof subset) |