[Test] Use flexible instance types for build_image compute queue

The cluster-creation step only submits 'sbatch -N 2 --wrap=uptime' and
makes no GPU/nvidia runtime assertions, so the queue does not need to be
pinned to the single tested (GPU) instance type. Render the ComputeResource
over flexible_instance_types, which the harness derives from the tested
instance plus architecture-matched siblings (same arch, vCPUs, threads,
EFA, and GPU presence via get_similar_instance_types). This further widens
capacity fallback beyond the multi-subnet change to reduce ICE failures
(V2280793096) while still satisfying the flexible-instance validators.

Author: himani2411

tests/integration-tests/tests/createami/test_createami/test_build_image/pcluster.config.yaml

@@ -21,4 +21,6 @@ Scheduling:
       ComputeResources:
         - Name: compute-resource1
           Instances:
-            - InstanceType: {{ instance }}
+            {% for instance_type in flexible_instance_types %}
+            - InstanceType: {{ instance_type }}
+            {% endfor %}