Rescue and report staging errors with telemetry

Staging errors from queue contention or other database issues would
cause the top level stager process to crash. Eventually that could shut
down the entire Oban supervision tree. Now we rescue standard database
connectivity issues instead and report them as errors in telemetry.

lib/oban/stager.ex

@@ -135,6 +135,8 @@ defmodule Oban.Stager do
 
       %{staged_count: length(staged), staged_jobs: staged}
     end)
+  rescue
+    error in [DBConnection.ConnectionError, Postgrex.Error] -> {:error, error}
   end
 
   defp stage_scheduled(state, leader?: true) do