feat(copilot): increase long-running operation TTL to 1 hour

- Increase long_running_operation_ttl from 600s (10min) to 3600s (1hour)
- Match stream_ttl duration for consistency
- Add clarifying description about deduplication lock purpose

Some operations (like complex agent runs) can take longer than 10 minutes.
The stream_registry heartbeat (publish_chunk) already keeps operations alive,
so this TTL is just a safety net for deduplication.

autogpt_platform/backend/backend/copilot/config.py

@@ -38,8 +38,10 @@ class ChatConfig(BaseSettings):
 
     # Long-running operation configuration
     long_running_operation_ttl: int = Field(
-        default=600,
-        description="TTL in seconds for long-running operation tracking in Redis (safety net if pod dies)",
+        default=3600,
+        description="TTL in seconds for long-running operation deduplication lock "
+        "(1 hour, matches stream_ttl). Prevents duplicate operations if pod dies. "
+        "For longer operations, the stream_registry heartbeat keeps them alive.",
     )
 
     # Stream registry configuration for SSE reconnection