Implement security hardening, notification pipeline, and CI test suite

Security hardening:

- Added IP/user rate limiting with Redis-backed counters and fail-open behavior.

- Added message anti-spam controls (per-chat rate + duplicate cooldown).

- Implemented refresh token rotation with JTI tracking and revoke support.

Notification pipeline:

- Added Celery app and async notification tasks for mention/offline delivery.

- Added Redis-based presence tracking and integrated it into realtime connect/disconnect.

- Added notification dispatch from message flow and notifications listing endpoint.

Quality gates and CI:

- Added pytest async integration tests for auth and chat/message lifecycle.

- Added pytest config, test fixtures, and GitHub Actions CI workflow.

- Fixed bcrypt/passlib compatibility by pinning bcrypt version.

- Documented worker and quality-gate commands in README.

app/notifications/schemas.py

@@ -1,7 +1,27 @@
-from pydantic import BaseModel
+from datetime import datetime
+from typing import Any
+
+from pydantic import BaseModel, ConfigDict
 
 
 class NotificationRequest(BaseModel):
     user_id: int
     event_type: str
     payload: dict
+
+
+class NotificationRead(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: int
+    user_id: int
+    event_type: str
+    payload: str
+    created_at: datetime
+
+
+class PushTaskPayload(BaseModel):
+    user_id: int
+    title: str
+    body: str
+    data: dict[str, Any]