[Model Monitoring] Fix default sampling_percentage on serving deplo…

…yment when monitoring is disabled (mlrun#7053)

When a user deploys a serving func without mm, the model endpoint is
generated with `monitoring mode` set to `disabled`. However, in this
case, the serving func has no `sampling_percentage` configuration, which
led to a bug during mep creation. This PR fixes this issue by setting a
default `sampling_percentage` value of 100 when it is not specified in
the serving spec.

https://iguazio.atlassian.net/browse/ML-8997

server/py/services/api/crud/model_monitoring/deployment.py

@@ -1309,9 +1309,9 @@ async def create_model_endpoints(self, function: dict, function_name: str):
             function_tag=function.metadata.tag,
             track_models=function.spec.track_models,
             graph=function.spec.graph,
-            sampling_percentage=function.spec.parameters[
-                mm_constants.EventFieldType.SAMPLING_PERCENTAGE
-            ],
+            sampling_percentage=function.spec.parameters.get(
+                mm_constants.EventFieldType.SAMPLING_PERCENTAGE, 100
+            ),
         )  # model endpoint, creation strategy, model path
         router_model_endpoints_instructions: list[
             tuple[