Attempt to remove flaky test for checking that tracer threads are cor…

…rectly shutdown (#3228)

Attempt to fix flaky test for checking that tracer threads are
correctly shutdown by adding back call to 
Timeout.ensure_timeout_thread_created that was removed in #2991

spec/datadog/integration_spec.rb

@@ -36,9 +36,11 @@ def inspect_threads(threads)
 
         shutdown
 
-        expect(threads.size).to eq(original_threads.size),
-          "Expected #{original_threads.size} threads (#{original_threads_inspect}), "\
-                  "got #{threads.size} threads (#{inspect_threads(threads)})"
+        post_shutdown_new_threads = Thread.list - original_threads
+
+        expect(post_shutdown_new_threads).to be_empty,
+          "Expected #{original_threads.size} threads but got #{threads.size} threads. "\
+            "extra_threads: (#{inspect_threads(post_shutdown_new_threads)})"
       end
     end
 

spec/spec_helper.rb

@@ -280,3 +280,11 @@ def initialize(*args, &block)
 
 # Helper matchers
 RSpec::Matchers.define_negated_matcher :not_be, :be
+
+# The Ruby Timeout class uses a long-lived class-level thread that is never terminated.
+# Creating it early here ensures tests that tests that check for leaking threads are not
+# triggered by the creation of this thread.
+#
+# This has to be one once for the lifetime of this process, and was introduced in Ruby 3.1.
+# Before 3.1, a thread was created and destroyed on every Timeout#timeout call.
+Timeout.ensure_timeout_thread_created if Timeout.respond_to?(:ensure_timeout_thread_created)