job/presubmit/ccm-aws: bump mem and cpu limit to prevent OOMKill

The idea of this PR is to bump resource utilization of e2e targeting
stability of existing presubmits which is curently having high falure[1]
ratio with many hours to get the feedback to the user[2].

Setting 3GiB/core to increase stability frm OOM kills.

[1]
The root cause of mostly failures cuased by CI infra is pointing to be
OOMKill. Here is one example of a e2e job using above mem and CPU limits:
https://monitoring-eks.prow.k8s.io/d/96Q8oOOZk/builds?orgId=1&var-org=kubernetes&var-repo=cloud-provider-aws&var-job=pull-cloud-provider-aws-e2e&var-build=All&from=1754491871179&to=1754494399603
https://issues.redhat.com/secure/attachment/13469904/13469904_Screenshot+From+2025-08-06+21-06-13.png
https://prow.k8s.io/view/gs/kubernetes-ci-logs/pr-logs/pull/cloud-provider-aws/1158/pull-cloud-provider-aws-e2e/1953110200760143872
https://kubernetes.slack.com/archives/C7J9RP96G/p1754505741634999

You can see instability on e2e presubmits recently (almost two weeks):
https://prow.k8s.io/job-history/gs/kubernetes-ci-logs/pr-logs/directory/pull-cloud-provider-aws-e2e

[2]  kubernetes-sigs/prow#210

Author: mtulio

config/jobs/kubernetes/cloud-provider-aws/cloud-provider-aws-presubmit.yaml

@@ -3,6 +3,8 @@ presubmits:
   - name: pull-cloud-provider-aws-e2e
     cluster: eks-prow-build-cluster
     decorate: true
+    decoration_config:
+      timeout: 4h
     skip_if_only_changed: "^docs/|^\\.github/|\\.(md)$|^(LICENSE|OWNERS)$"
     skip_branches:
     - gh-pages
@@ -17,10 +19,10 @@ presubmits:
         resources:
           limits:
             cpu: 2
-            memory: 4Gi
+            memory: 6Gi
           requests:
             cpu: 2
-            memory: 4Gi
+            memory: 6Gi
         command:
         - runner.sh
         args: