Add offline unit test suite for chat_backend (#5)

Replaces the three scattered test modules with a `chat_backend/tests/` package
covering models, DatabaseStorage, serializers, every REST endpoint, the LLM
services, signals and both websocket consumers — 242 deterministic tests that
run without Ollama, Chroma, SMTP or network access, plus 6 opt-in live-Ollama
checks behind RUN_LIVE_OLLAMA_TESTS.

A custom test runner sets SKIP_RAG_INIT and a cheap password hasher so the suite
finishes in seconds and can no longer reach a model server by accident.

Bugs the new tests surfaced, fixed here:
- views.ConversationDetailView.post silently dropped every prompt: `import
  datetime` shadowed `from datetime import datetime`, so `datetime.now()` raised
  inside a bare except.
- get_conversation_file_async in both consumers always returned None with
  DatabaseStorage, because `prompt.file.read` opens the blob (a DB query) during
  attribute access in async context.
- DatabaseStorage._save crashed on str-backed ContentFile.
- prompt_classifier/__init__.,py was never importable as a package init.

Co-authored-by: Cursor <cursoragent@cursor.com>

README.md

@@ -44,11 +44,24 @@ uv run python manage.py runserver 0.0.0.0:8003
 
 Without `DATABASE_URL` / `DB_HOST`, settings fall back to SQLite (`llm_be/db.sqlite3`).
 
-Tests (skip live-Ollama classifier cases):
+Tests:
 
 ```bash
 cd llm_be
-SKIP_RAG_INIT=1 uv run python manage.py test
+uv run python manage.py test
+```
+
+The suite is offline by default — the custom test runner
+(`llm_be/test_runner.py`) sets `SKIP_RAG_INIT=1` and a cheap password hasher, and
+LLM chains are faked, so no Ollama, Chroma, SMTP or network access is needed.
+Tests live in `llm_be/chat_backend/tests/` (models, storage, serializers, views,
+services, signals, consumers).
+
+Non-deterministic checks against a real model server are opt-in:
+
+```bash
+cd llm_be
+RUN_LIVE_OLLAMA_TESTS=1 uv run python manage.py test chat_backend.tests.test_live_ollama
 ```
 
 ### Docker (dev, bundled Postgres)