[OCTRL-1064] GO_ERROR calls should never be critical

Having to handle a failed GO_ERROR transition is something we should avoid. An interrupted GO_ERROR means that we might not perform the complete procedure and miss e.g. calling some services/subsystems.

With this change, the only known way of failing GO_ERROR is to try it in an inappropriate state, which can be ERROR (no point in transitioning from ERROR to ERROR) or DONE (too late, nothing to do).

Author: knopers8

workflows/readout-dataflow.yaml

@@ -1946,7 +1946,7 @@ roles:
           trigger: before_GO_ERROR
           await: after_GO_ERROR-1
           timeout: "{{ odc_goerror_timeout }}"
-          critical: true
+          critical: false
       - name: reset
         call:
           func: odc.Reset()