Add LLM key verification steps to integration_tests and e2e_tests CI
jobs. When Anthropic/OpenAI/Google API keys are not configured, these
jobs skip gracefully with an informative message instead of failing.

Update status-check to treat skipped results from integration_tests and
e2e_tests as acceptable (only fail on failure/cancelled), while keeping
strict success-only requirements for core jobs (lint, typecheck, security,
quality, unit_tests, coverage, build, docker, helm, push-validation).

Add BDD tests verifying the guard mechanism exists and behaves correctly.

ISSUES CLOSED: #9222

@@ -269,0 +296,4 @@
+                      echo "Skipping integration_tests job."
+                      exit 0
+                  fi
+

@@ -608,7 +676,7 @@ jobs:
                   fi

@@ -648,0 +716,4 @@
+                  # LLM-dependent jobs (integration_tests, e2e_tests) may skip
+                  # gracefully when secrets are unavailable. Accept success or skip;
+                  # only fail on failure or cancelled.
+                  for job in integration_tests e2e_tests; do

@@ -10,0 +21,4 @@
+  merging when only non-LLM aspects of the change were modified. The `benchmark-regression`
+  job in `master.yml` already used this pattern; this brings consistency to all jobs that
+  gate on external secrets (see lines 98-110 of `master.yml`).
+

@@ -0,0 +134,4 @@
+@then("its status-check result logic should handle \"skipped\" as acceptable")
+def step_then_status_check_handles_skipped(context):
+    """Verify that the status-check job treats skipped/success as acceptable for LLM-dependent jobs."""
+    import re as _re

@@ -0,0 +14,4 @@
+    And it should have a step that runs shell code containing "OPENAI_API_KEY"
+
+  Scenario: e2e_tests job has LLM key verification step
+    When I parse the CI workflow YAML

@@ -0,0 +138,4 @@
+
+    jobs = context.ci_workflow.get("jobs", {})
+    status_check = jobs.get("status-check")
+    if status_check is None: