Fix race condition when terminating supervisor after restoring default signals

We'd terminate the supervisor and let it consume the shutdown timeout, so it'd start
the immediate shutdown, then do some assertions and terminate the supervisor again
before it had the chance to finish, and as such, the TERM signal would arrive sometimes
before and sometimes after default signal handlers would have been restored, resulting in
an error (harmless, but annoying) as:
```
..solid_queue/test/test_helper.rb:51: warning: Exception in finalizer #<Proc:0x00000001083ee550 (lambda)>
..solid_queue/test/test_helper.rb:51:in `fork': SIGTERM (SignalException)
	from ..solid_queue/test/test_helper.rb:51:in `run_supervisor_as_fork'
	from ..solid_queue/test/integration/processes_lifecycle_test.rb:9:in `block in <class:ProcessLifecycleTest>'
```

Author: rosa

test/integration/processes_lifecycle_test.rb

@@ -117,7 +117,7 @@ class ProcessLifecycleTest < ActiveSupport::TestCase
 
   test "term supervisor exceeding timeout while there are jobs in-flight" do
     no_pause = enqueue_store_result_job("no pause")
-    pause = enqueue_store_result_job("pause", pause: SolidQueue.shutdown_timeout + 1.second)
+    pause = enqueue_store_result_job("pause", pause: SolidQueue.shutdown_timeout + 10.second)
 
     signal_process(@pid, :TERM, wait: 0.1.second)
     wait_for_jobs_to_finish_for(SolidQueue.shutdown_timeout + 0.1.second)
@@ -132,6 +132,10 @@ class ProcessLifecycleTest < ActiveSupport::TestCase
 
     # The process running the long job couldn't deregister, the other did
     assert_registered_workers_for(:background)
+
+    # Now wait until the supervisor finishes for real, which will complete the cleanup
+    wait_for_process_termination_with_timeout(@pid, timeout: 1.second)
+    assert_clean_termination
   end
 
   test "process some jobs that raise errors" do