[evaluation] add retry logic for high concurrency scenarios in AdversarailSimulator, _SafetyEvaluation (#41978)

* Add retry logic for high concurrency scenarios

* Update sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/simulator/_model_tools/_proxy_completion_model.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* black fixes

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Author: slister1001

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/simulator/_model_tools/_proxy_completion_model.py

@@ -10,7 +10,7 @@
 
 from azure.ai.evaluation._http_utils import AsyncHttpPipeline, get_async_http_client
 from azure.ai.evaluation._user_agent import UserAgentSingleton
-from azure.core.exceptions import HttpResponseError
+from azure.core.exceptions import HttpResponseError, ServiceResponseError
 from azure.core.pipeline.policies import AsyncRetryPolicy, RetryMode
 from azure.ai.evaluation._common.onedp._client import AIProjectClient
 from azure.ai.evaluation._common.onedp.models import SimulationDTO
@@ -225,7 +225,25 @@ async def request_api(
                     sleep_time = RAIService.SLEEP_TIME**request_count
                     await asyncio.sleep(sleep_time)
         else:
-            response = await session.post(url=self.endpoint_url, headers=proxy_headers, json=sim_request_dto.to_dict())
+            # Retry policy for POST request to RAI service
+            service_call_retry_policy = AsyncRetryPolicy(
+                retry_on_exceptions=[ServiceResponseError],
+                retry_total=7,
+                retry_backoff_factor=10.0,
+                retry_backoff_max=180,
+                retry_mode=RetryMode.Exponential,
+            )
+
+            response = None
+            async with get_async_http_client().with_policies(retry_policy=service_call_retry_policy) as retry_client:
+                try:
+                    response = await retry_client.post(
+                        url=self.endpoint_url, headers=proxy_headers, json=sim_request_dto.to_dict()
+                    )
+                except ServiceResponseError as e:
+                    self.logger.error("ServiceResponseError during POST request to rai svc after retries: %s", str(e))
+                    raise
+
             # response.raise_for_status()
             if response.status_code != 202:
                 raise HttpResponseError(