test: fix a test crash by destroying test time after consumers

[alyssawilk]

also pretty printing grpc test params while I'm in there because I despise figuring out what test is failing via staring at Test/0 Test/1 Test/2 Test/3

Risk Level: Low (test only)
Testing: goes from 80% failure with flake-options on to 0% failure
Docs Changes: n/a
Release Notes: n/a
Fixes #5071

[mattklein123]

nice!

[adityagujral]

Hi,
could you please explain how this commit fixes the issue. I am facing a similar issue and I am curious to understand the fix.

[alyssawilk]

I'm pulling from memory but I think basically some other objects (fake upstream?) would occasionally and non-deterministically use fake time in its destructor (time based because it happened if the stream wasn't fully cleaned up). Because the time object was deleted before the object referencing it, it caused problems.

[adityagujral]

Thanks. But how does your change in grpc_client_integration_test_harness.h address this exactly? Sorry if I am missing something very obvious as I'm new to cpp. I am seeing a similar issue in a new integration test only when using GCC and not in Clang.

[alyssawilk]

in C++ class variables are destroyed bottom up, so making test_time the first in the test made sure it'd be destroyed last (once all other members were done using it)