fix(insights): don't hold DB transaction across LLM API calls

The Repo.transaction wrapping Repo.stream() held a checkout for the
entire iteration, including the LLM HTTP call. When the LLM took >15s,
Postgrex killed the connection, cascading to all subsequent orgs.
Collect org IDs first, then process each independently.

lib/towerops/workers/ai_network_insight_worker.ex

@@ -32,19 +32,14 @@ defmodule Towerops.Workers.AiNetworkInsightWorker do
 
   @impl Oban.Worker
   def perform(%Oban.Job{}) do
-    case Repo.transaction(fn ->
-           Organization
-           |> select([o], o.id)
-           |> Repo.stream()
-           |> Enum.each(&process_organization/1)
-         end) do
-      {:ok, _} ->
-        :ok
+    org_ids =
+      Organization
+      |> select([o], o.id)
+      |> Repo.all()
 
-      {:error, reason} ->
-        Logger.error("AI network insight transaction failed: #{inspect(reason)}")
-        {:error, reason}
-    end
+    Enum.each(org_ids, &process_organization/1)
+
+    :ok
   end
 
   defp process_organization(organization_id) do