fix(consumer): Ensure backpressure is properly handled in stale message routing (#83271)

Previously we were not handling the case where a `MessageRejected` error
which signals backpressure from downstream steps might be received in
the dlq stale messages strategy. If this happens we can just ignore the
error, we will retry on the next loop.

Author: lynnagara

src/sentry/consumers/dlq.py

@@ -7,7 +7,11 @@
 
 from arroyo.backends.kafka import KafkaPayload, KafkaProducer
 from arroyo.dlq import InvalidMessage, KafkaDlqProducer
-from arroyo.processing.strategies.abstract import ProcessingStrategy, ProcessingStrategyFactory
+from arroyo.processing.strategies.abstract import (
+    MessageRejected,
+    ProcessingStrategy,
+    ProcessingStrategyFactory,
+)
 from arroyo.types import FILTERED_PAYLOAD, BrokerValue, Commit, FilteredPayload, Message, Partition
 from arroyo.types import Topic as ArroyoTopic
 from arroyo.types import Value
@@ -127,9 +131,12 @@ def poll(self) -> None:
         if self.offsets_to_forward:
             if time.time() > self.last_forwarded_offsets + 1:
                 filtered_message = Message(Value(FILTERED_PAYLOAD, self.offsets_to_forward))
-                self.next_step.submit(filtered_message)
-                self.offsets_to_forward = {}
-                self.last_forwarded_offsets = time.time()
+                try:
+                    self.next_step.submit(filtered_message)
+                    self.offsets_to_forward = {}
+                    self.last_forwarded_offsets = time.time()
+                except MessageRejected:
+                    pass
 
     def join(self, timeout: float | None = None) -> None:
         self.next_step.join(timeout)