Fix a crash when an Executor wrapped fork exit.

Fix rails#51298

forking inside an Execution Wrapper crashes when running the completion callbacks.

Rails 7.0 was not Execution wrapping the `runner` command.
Rails 7.2 changed the definition of `active_connection?`, the new definition doesn't contain the bug.

Therefore forking inside a script intended to be run with the `runner` command on 7.1 crashes. (see rails#51298)

activerecord/lib/active_record/query_cache.rb

@@ -30,7 +30,7 @@ def self.run
     end
 
     def self.complete(pools)
-      pools.each { |pool| pool.disable_query_cache! }
+      pools.each { |pool| pool.disable_query_cache! unless pool.discarded? }
 
       ActiveRecord::Base.connection_handler.each_connection_pool do |pool|
         pool.release_connection if pool.active_connection? && !pool.connection.transaction_open?

activerecord/test/cases/query_cache_test.rb

@@ -650,6 +650,37 @@ def test_clear_query_cache_is_called_on_all_connections
     ActiveRecord::Base.connection_pool.lock_thread = false
   end
 
+  test "query cache callbacks exit gracefully from a fork" do
+    # Regression test for #51298
+    pid = nil
+    status = nil
+
+    # Run forking Rack app
+    middleware { |env|
+      pid = fork
+      _, status = Process.wait2(pid) if pid.present?
+    [200, {}, nil]
+    }.call({})
+
+
+    # Happy path, the fork exit 0 and the main process
+    # asserts on the exit status of the fork
+    if pid.nil?
+      exit 0
+    else
+      assert_equal 0, status.exitstatus
+    end
+  rescue
+    # Unhappy path, the fork exit 1.
+    # The main process should not reach here, but in case
+    # something's wrong we need to re-raise the error.
+    if pid.nil?
+      exit 1
+    else
+      raise
+    end
+  end
+
   private
     def with_temporary_connection_pool(&block)
       pool_config = ActiveRecord::Base.connection.pool.pool_config