Work around bug that was breaking q.empty()

[relud]

peter-wangxu/persist-queue#154 can cause q.empty() to return inaccurate results, which in turn can cause the flush manager to incorrectly think that disks were flushed because this loop wouldn't drain the queue:

gcp-ingestion/ingestion-edge/ingestion_edge/flush.py

Line 134 in 70b8960

while not self.q.empty():

[whd]

This does appear to fix the error case we're debugging, but tests are failing and the upstream fix PR is also failing CI.

[whd]

I'm slightly worried that we're catching this now instead of in a test case somewhere, but LGTM

[relud]

I'm slightly worried that we're catching this now instead of in a test case somewhere, but LGTM

filed #1635 to track adding a relevant test