Bugfix/do not overwrite stable at (#4301)

* Fixes race condition where the updated incident is marked as stable, but the update flow has not yet completed.

* Reduces likelihood of incident_update_flow() running before incident_service.update() completes.

* Adds a check to only set the stable at field if the incident is not active and the stable_at field has not been set yet.

* Restores previous version of incident update functions

src/dispatch/incident_cost/service.py

@@ -225,7 +225,8 @@ def calculate_incident_response_cost_with_classic_model(incident: Incident, inci
                     participant_role_renounced_at = participant_role.renounced_at
             else:
                 # Fixes race condition where the updated incident is marked as stable, but the update flow has not yet completed.
-                incident.stable_at = datetime.utcnow()
+                if not incident.stable_at:
+                    incident.stable_at = datetime.utcnow()
 
                 # we set the renounced_at default time to the stable_at time
                 participant_role_renounced_at = incident.stable_at